fix(stats): preserve manual episode assignments across reparsing

The reassignAnimeAnilist redistribution test seeded raw sessions and
telemetry but not the imm_lifetime_media rows a finalized session leaves
behind. The old destructive lifetime rebuild re-derived summaries from
raw sessions and hid that gap; the non-destructive recompute reads
per-video lifetime rows, so the fixture must contain them.

CHANGELOG.md

@@ -1,5 +1,56 @@
 # Changelog
 
+## v0.19.3 (2026-08-13)
+
+### Added
+- Changelog Modal: Adds an in-app changelog you can open from the tray ("View Changelog") or the "What's New" button on the update notification, so the notification stays reachable while you read. It shows the newest published release notes (falling back to the bundled changelog if that fetch fails), folds older versions while keeping the current one expanded, and supports keyboard navigation (`J`/`K`/arrows, `Enter`, `R`, `Esc`).
+
+### Changed
+- Subtitle Tokenization Performance: Reworks subtitle dictionary lookups to cut per-line work roughly in half, cache repeated lookups across lines, and stop tokenization from competing with on-screen subtitle prefetching. Also fixes several accuracy issues along the way: dropped readings on trailing kana, character names being skipped after a dictionary sync, annotations not refreshing after mining a card, and halfwidth katakana character names losing their reading or being swallowed by other words.
+
+### Fixed
+- Character Dictionary Large Imports: Large character dictionaries (e.g. One Piece) no longer fail to install from a fixed timeout budget; the import now scales its time budget to dictionary size and reports detailed progress (page/character counts, image download progress, elapsed time) instead of one static message.
+- Stats Delete Responsiveness: Deleting sessions, episodes, or library entries no longer freezes the stats page or an active video player; deletes are now batched into a single transaction.
+- Styled Subtitle Cue Parsing: Heavily typeset subtitles (karaoke, signs) no longer flood the subtitle sidebar with garbage; vector drawing commands are no longer shown as text, and duplicate/animation-burst cues now collapse into one.
+- X11 mpv Renderer: Fixes an mpv crash on the first fullscreen toggle for X11/XWayland users with `gpu-next` shaders (e.g. ArtCNN), which was caused by X11 mode forcing the legacy OpenGL renderer.
+- X11 Overlay Display Scaling: Fixes the overlay appearing oversized and offset from mpv on X11/XWayland under fractional or mixed-monitor display scaling.
+
+<details>
+<summary>Internal changes</summary>
+
+### Internal
+- Subtitle text is now decoded from ASS exactly once at ingest, so the renderer, timing tracker, and tokenizer all share one decoded value instead of each re-deriving it.
+- Added per-stage debug timings (`scanMs`, `mecabMs`, `frequencyMs`, `annotateMs`) to the subtitle tokenization pipeline log.
+
+</details>
+
+## v0.19.2 (2026-08-04)
+
+### Changed
+- Subsync: The sync modal now lets you choose both the reference subtitle (correct timing) and the out-of-sync subtitle to retime, for both alass and ffsubsync. alass can also use the loaded video's audio as a reference for local files. Retiming the secondary track now reloads the result into the secondary slot instead of overwriting the primary subtitle.
+
+### Fixed
+- Streaming Subtitle Tokenization: Jellyfin streams now seed subtitle tokenization directly from the downloaded subtitle file instead of relying on an mpv event that could be missed, and prefetching now runs to the end of the file and clears between episodes. The tokenization cache was raised from 256 to 2500 lines, and parsed cues are no longer lost when the active subtitle track briefly can't be resolved (e.g. switching to an embedded track). Together these prevent episodes from falling back to slow, line-by-line tokenization during playback.
+- Overlay: Subtitle lines now appear immediately at their cue time even on a tokenization cache miss, upgrading in place once tokens and annotations are ready, instead of waiting on a line still being processed. A failed tokenization is no longer cached as plain text, so repeated lines get another chance at annotations.
+- Background Logging: Background startup now respects the configured logging level when no explicit log level is passed.
+
+<details>
+<summary>Internal changes</summary>
+
+### Internal
+- Patched three high-severity dependency advisories (`undici`, `brace-expansion`, `fast-uri`).
+
+</details>
+
+## v0.19.1 (2026-08-01)
+
+### Added
+- Word Card Type: Adds a setting (Settings > Mining/Anki > Kiku/Lapis Features > "Word Card Type") to choose which card-type flag SubMiner marks on Kiku/Lapis word cards — `word-and-sentence` (default), `click`, `sentence`, `audio`, or `none`. Click cards (`IsClickCard`) can now be flagged, and setting any card-type flag clears the others so a note can't claim two types at once.
+
+### Fixed
+- Yomitan Popup: Fixes the macOS Yomitan popup going inert after mining a card — clicks outside the popup no longer pass through to mpv, and scrolling over the popup scrolls its definitions instead of seeking playback.
+- YouTube Playlist Links: Fixes opening a video from a playlist URL (e.g. a Watch Later link with `list=`/`index=`) timing out while probing subtitles, metadata, or the playback URL.
+
 ## v0.19.0 (2026-07-29)
 
 ### Added

CLAUDE.md

@@ -0,0 +1 @@
+AGENTS.md

bun.lock

@@ -26,7 +26,7 @@
         "eslint": "^10.8.0",
         "prettier": "^3.8.1",
         "typescript": "^5.9.3",
-        "undici": "7.28.0",
+        "undici": "7.29.0",
       },
     },
   },
@@ -36,16 +36,18 @@
   "overrides": {
     "@xmldom/xmldom": "0.8.13",
     "app-builder-lib": "26.15.3",
-    "brace-expansion": "5.0.8",
+    "brace-expansion": "5.0.9",
     "electron-builder-squirrel-windows": "26.15.3",
+    "fast-uri": "3.1.5",
     "form-data": "4.0.6",
     "ip-address": "10.2.0",
-    "js-yaml": "4.3.0",
+    "js-yaml": "4.3.1",
     "lodash": "4.18.0",
     "minimatch": "10.2.5",
     "picomatch": "4.0.4",
     "tar": "7.5.21",
     "tmp": "0.2.7",
+    "undici": "7.29.0",
   },
   "packages": {
     "@discordjs/collection": ["@discordjs/collection@2.1.1", "", {}, "sha512-LiSusze9Tc7qF03sLCujF5iZp7K+vRNEDBZ86FT9aQAv3vxMLihUvKvpsCWiQ2DJq1tVckopKm1rxomgNUc9hg=="],
@@ -266,7 +268,7 @@
 
     "boolean": ["boolean@3.2.0", "", {}, "sha512-d0II/GO9uf9lfUHH2BQsjxzRJZBdsjgsBiW4BvhWk/3qoKwQFjIDVN19PfX8F2D/r9PCMTtLWjYVCFrpeYUzsw=="],
 
-    "brace-expansion": ["brace-expansion@5.0.8", "", { "dependencies": { "balanced-match": "^4.0.2" } }, "sha512-JZyDyq3D4AUifKTPOB7DELf6XsB3WdPuNxCtob1vFXPsSXhdAiHBWJ/tJ8HAc9aH84BK+5JFZLNkJKx3G9kzQg=="],
+    "brace-expansion": ["brace-expansion@5.0.9", "", { "dependencies": { "balanced-match": "^4.0.2" } }, "sha512-ScQ4IuvIEF1TMlP7Zt+vjJ//9zlPb2SDcxWxM3bk8s6t6GGdJ7KO1dCcTidOPJKePW30LE/2cT7wCyPho9/Wxg=="],
 
     "buffer-from": ["buffer-from@1.1.2", "", {}, "sha512-E+XQCRwSbaaiChtv6k6Dwgc+bx+Bs6vuKJHHl5kox/BaKbhiXzqQOwK4cO22yElGp2OCmjwVhT3HmxgyPGnJfQ=="],
 
@@ -404,7 +406,7 @@
 
     "fast-levenshtein": ["fast-levenshtein@2.0.6", "", {}, "sha512-DCXu6Ifhqcks7TZKY3Hxp3y6qphY5SJZmrWMDrKcERSOXWQdMhU9Ig/PYrzyw/ul9jOIyh0N4M0tbC5hodg8dw=="],
 
-    "fast-uri": ["fast-uri@3.1.4", "", {}, "sha512-8JnbkQ4juDyvYs4mgFGQqg4yCYtFDtUtmp2QIQq11ZZe5CFQ5wcqm1rqDgAh/QdMySuBnPzMUiJUNZG5N/AiQw=="],
+    "fast-uri": ["fast-uri@3.1.5", "", {}, "sha512-gHwA1O9LDIcKunMKhObS/HimwtehO1nPUECKAu5TpKgaO19fcWEl4bliWe1jWxVFvIXztJjjQ4L8XQ1EU9f7Jw=="],
 
     "fdir": ["fdir@6.5.0", "", { "peerDependencies": { "picomatch": "^3 || ^4" }, "optionalPeers": ["picomatch"] }, "sha512-tIbYtZbucOs0BRGqPJkshJUYdL+SDH7dVM8gjy+ERp3WAUjLEFJE+02kanyHtwjWOnwrKYBiwAmM0p4kLJAnXg=="],
 
@@ -496,7 +498,7 @@
 
     "jiti": ["jiti@2.6.1", "", { "bin": { "jiti": "lib/jiti-cli.mjs" } }, "sha512-ekilCSN1jwRvIbgeg/57YFh8qQDNbwDb9xT/qu2DAHbFFZUicIl4ygVaAvzveMhMVr3LnpSKTNnwt8PoOfmKhQ=="],
 
-    "js-yaml": ["js-yaml@4.3.0", "", { "dependencies": { "argparse": "^2.0.1" }, "bin": { "js-yaml": "bin/js-yaml.js" } }, "sha512-1td788aAnnZ5qs7V2QIRl1owjtYpbKt749Y3xauqQgwIIGF/xXWz1wMTEBx5O3LK3lXLVuqXPdPxj2BoFHaW9Q=="],
+    "js-yaml": ["js-yaml@4.3.1", "", { "dependencies": { "argparse": "^2.0.1" }, "bin": { "js-yaml": "bin/js-yaml.js" } }, "sha512-CY6crGq313MX8GkwvB7tzgp99vjQxY1++5y10/BKN/GUfHqWaOGQMNZkBvqSzsZKWk/ijwHlWzzkLulsGHhjWQ=="],
 
     "json-buffer": ["json-buffer@3.0.1", "", {}, "sha512-4bV5BfR2mqfQTJm+V5tPPdf+ZpuhiIvTuAB5g8kcrXOZpTT/QwwVRWBywX1ozr6lEuPdbHxwaJlm9G6mI2sfSQ=="],
 
@@ -714,7 +716,7 @@
 
     "typescript": ["typescript@5.9.3", "", { "bin": { "tsc": "bin/tsc", "tsserver": "bin/tsserver" } }, "sha512-jl1vZzPDinLr9eUt3J/t7V6FgNEw9QjvBPdysz9KfQDD41fQrC2Y4vKQdiaUpFT4bXlb1RHhLpp8wtm6M5TgSw=="],
 
-    "undici": ["undici@7.28.0", "", {}, "sha512-cRZYrTDwWznlnRiPjggAGxZXanty6M8RV1ff8Wm4LWXBp7/IG8v5DnOm74DtUBp9OONpK75YlPnIjQqX0dBDtA=="],
+    "undici": ["undici@7.29.0", "", {}, "sha512-IDxfleLmmbSskfWSUATiN1nfn2rDuvnMOqb5CWR92iIfojA0Ud+ulOAAEQ57LPr9rWmsreUyf5lwyao+7GNNVw=="],
 
     "undici-types": ["undici-types@7.16.0", "", {}, "sha512-Zz+aZWSj8LE6zoxD+xrjh4VfkIG8Ya6LvYkZqtUQGJPZjYl53ypCaUwWqo7eI0x66KBGeRo+mlBEkMSeSZ38Nw=="],
 
@@ -754,8 +756,6 @@
 
     "@discordjs/rest/discord-api-types": ["discord-api-types@0.38.43", "", {}, "sha512-sSoBf/nK6m7BGtw65mi+QBuvEWaHE8MMziFLqWL+gT6ME/BLg34dRSVKS3Husx40uU06bvxUc3/X+D9Y6/zAbw=="],
 
-    "@discordjs/rest/undici": ["undici@6.27.0", "", {}, "sha512-YmfV3YnEDzXRC5lZ2jWtWWHKGUm1zIt8AhesR1tens+HTNv+YZlN/dp6G727LOvMJ8xjP9Be7Y2Sdr96LDm+pg=="],
-
     "@discordjs/util/discord-api-types": ["discord-api-types@0.38.43", "", {}, "sha512-sSoBf/nK6m7BGtw65mi+QBuvEWaHE8MMziFLqWL+gT6ME/BLg34dRSVKS3Husx40uU06bvxUc3/X+D9Y6/zAbw=="],
 
     "@electron/asar/commander": ["commander@5.1.0", "", {}, "sha512-P0CysNDQ7rtVw4QIQtm+MRxV66vKFSvlsQvGYXZWR3qFU0jlMKHZZZgw8e+8DSah4UDKMqnknRDQz+xuQXQ/Zg=="],
@@ -808,8 +808,6 @@
 
     "node-gyp/env-paths": ["env-paths@2.2.1", "", {}, "sha512-+h1lkLKhZMTYjog1VEpJNG7NZJWcuc2DDk/qsqSTRRCOXiLjeQ1d1/udrUGhqMxUgAlwKNZ0cf2uqan5GLuS2A=="],
 
-    "node-gyp/undici": ["undici@6.27.0", "", {}, "sha512-YmfV3YnEDzXRC5lZ2jWtWWHKGUm1zIt8AhesR1tens+HTNv+YZlN/dp6G727LOvMJ8xjP9Be7Y2Sdr96LDm+pg=="],
-
     "node-gyp/which": ["which@6.0.1", "", { "dependencies": { "isexe": "^4.0.0" }, "bin": { "node-which": "bin/which.js" } }, "sha512-oGLe46MIrCRqX7ytPUf66EAYvdeMIZYn3WaocqqKZAxrBpkqHfL/qvTyJ/bTk5+AqHCjXmrv3CEWgy368zhRUg=="],
 
     "pkijs/@noble/hashes": ["@noble/hashes@1.4.0", "", {}, "sha512-V1JJ1WTRUqHHrOSh597hURcMqVKVGL/ea3kv0gSnEdsEZ0/+VyPghM1lMNGc00z7CIQorSvbKpuJkxvuHbvdbg=="],

changes/library-merge-and-move.md

@@ -0,0 +1,6 @@
+type: added
+area: stats
+
+- Library: duplicate cards for the same show can now be combined. Press "Select" above the library grid, tick the cards, and use "Merge Selected"; the dialog picks which entry to keep and moves every episode onto it. Sessions, mined cards, and watch time are preserved, the emptied entries disappear, and remembered title aliases keep future episodes on the merged card.
+- Library: episodes can be reassigned to another library entry from the "→" button on an episode row, which is the fix when one file lands under a stray title (e.g. an episode name parsed as the series). Manual assignments now survive later filename parsing, Jellyfin refreshes, and season repair. Compatible local episodes in the same directory reuse a uniquely corrected destination, while conflicting seasons or manual destinations are not forced together. Emptying an entry this way removes it and returns to the grid.
+- Library: exact AniList title matches with compatible seasons fold duplicate cards automatically. Fuzzy same-AniList matches appear as dismissible "Possible duplicate" reviews instead of changing the library without confirmation; conflicting explicit seasons are left alone.

config.example.jsonc

@@ -523,7 +523,7 @@
   // ==========================================
   // AnkiConnect Integration
   // Automatic Anki updates and media generation options.
-  // Hot-reload: ankiConnect.ai.enabled, media.normalizeAudio/mirrorMpvVolume, knownWords, nPlusOne, fields.word/audio/image/sentence/miscInfo, behavior.autoUpdateNewCards, isLapis.sentenceCardModel, and isKiku.fieldGrouping update live while SubMiner is running.
+  // Hot-reload: ankiConnect.ai.enabled, media.normalizeAudio/mirrorMpvVolume, knownWords, nPlusOne, fields.word/audio/image/sentence/miscInfo, behavior.autoUpdateNewCards, isLapis.sentenceCardModel, isKiku.fieldGrouping, and lapisKiku.wordCardKind update live while SubMiner is running.
   // Shared AI provider transport settings are read from top-level ai and typically require restart.
   // Most other AnkiConnect settings still require restart.
   // ==========================================
@@ -605,7 +605,10 @@
       "enabled": false, // Enable Kiku-specific mining behaviors (duplicate handling, field grouping). Values: true | false
       "fieldGrouping": "disabled", // Kiku duplicate-card field grouping mode. Values: auto | manual | disabled
       "deleteDuplicateInAuto": true // When Kiku field grouping is "auto", delete the duplicate source card after grouping completes. Values: true | false
-    } // Is kiku setting.
+    }, // Is kiku setting.
+    "lapisKiku": {
+      "wordCardKind": "word-and-sentence" // Card-type flag SubMiner marks on Kiku/Lapis word cards. Only one flag is set at a time; the others are cleared. Requires isKiku.enabled or isLapis.enabled. Values: word-and-sentence | click | sentence | audio | none
+    } // Lapis kiku setting.
   }, // Automatic Anki updates and media generation options.
 
   // ==========================================

docs-site/anki-integration.md

@@ -289,6 +289,21 @@ Trigger with the mine sentence shortcut (`Ctrl/Cmd+S` by default). The card is c
 
 To mine multiple subtitle lines as one sentence card, use `Ctrl/Cmd+Shift+S` followed by a digit (1–9) to select how many recent lines to combine.
 
+## Word Card Type (Kiku/Lapis)
+
+Word cards get a card-type flag when SubMiner fills their sentence, whether that comes from Yomitan auto-enrichment, a manual clipboard update, or stats-dashboard word mining. By default the flag is `IsWordAndSentenceCard`; pick a different one with `ankiConnect.lapisKiku.wordCardKind`.
+
+```jsonc
+"ankiConnect": {
+  "isKiku": { "enabled": true },
+  "lapisKiku": {
+    "wordCardKind": "click" // word-and-sentence (default), click, sentence, audio, none
+  }
+}
+```
+
+`click` marks `IsClickCard`, `sentence` marks `IsSentenceCard`, `audio` marks `IsAudioCard`, and `none` leaves the flags untouched for templates that manage them elsewhere. Whichever flag is chosen, the other card-type flags are cleared so the note never claims two card types. The setting is only read when `isKiku` or `isLapis` is enabled, and cards mined with Mine Sentence or Mine Audio keep their own flag.
+
 ## Field Grouping (Kiku)
 
 When you mine the same word multiple times, SubMiner can merge the cards instead of creating duplicates. This is designed for note types like [Kiku](https://github.com/youyoumu/kiku) that support grouped sentence/audio/image fields.
@@ -313,11 +328,11 @@ When you mine the same word multiple times, SubMiner can merge the cards instead
 
 ### What Gets Merged
 
-| Field    | Merge behavior                          |
+| Field    | Merge behavior                                |
-| -------- | ---------------------------------------- |
+| -------- | --------------------------------------------- |
 | Sentence | Both cards' sentences kept as grouped entries |
-| Audio    | Both cards' `[sound:...]` entries kept  |
+| Audio    | Both cards' `[sound:...]` entries kept        |
-| Image    | Both cards' images kept                 |
+| Image    | Both cards' images kept                       |
 
 Identical values from both cards are kept as separate grouped entries; the merge does not deduplicate.
 

docs-site/architecture.md

@@ -75,8 +75,8 @@ src/
   renderer/                # Overlay renderer (modularized UI/runtime)
     handlers/              # Keyboard/mouse/gamepad interaction modules
     modals/                # Modal flows (Jimaku, Kiku, subsync, runtime options, session help,
-                           #   character dictionary, playlist browser, subtitle sidebar,
+                           #   changelog, character dictionary, playlist browser, subtitle
-                           #   YouTube track picker, controller config/debug/select)
+                           #   sidebar, YouTube track picker, controller config/debug/select)
     positioning/           # Subtitle position controller (drag-to-reposition)
   settings/                # Settings window UI (model, controls, markup)
   types/                   # Domain type modules (anki, config, integrations, ...)

docs-site/changelog.md

@@ -1,5 +1,56 @@
 # Changelog
 
+## v0.19.3 (2026-08-13)
+
+**Added**
+- Changelog Modal: Adds an in-app changelog you can open from the tray ("View Changelog") or the "What's New" button on the update notification, so the notification stays reachable while you read. It shows the newest published release notes (falling back to the bundled changelog if that fetch fails), folds older versions while keeping the current one expanded, and supports keyboard navigation (`J`/`K`/arrows, `Enter`, `R`, `Esc`).
+
+**Changed**
+- Subtitle Tokenization Performance: Reworks subtitle dictionary lookups to cut per-line work roughly in half, cache repeated lookups across lines, and stop tokenization from competing with on-screen subtitle prefetching. Also fixes several accuracy issues along the way: dropped readings on trailing kana, character names being skipped after a dictionary sync, annotations not refreshing after mining a card, and halfwidth katakana character names losing their reading or being swallowed by other words.
+
+**Fixed**
+- Character Dictionary Large Imports: Large character dictionaries (e.g. One Piece) no longer fail to install from a fixed timeout budget; the import now scales its time budget to dictionary size and reports detailed progress (page/character counts, image download progress, elapsed time) instead of one static message.
+- Stats Delete Responsiveness: Deleting sessions, episodes, or library entries no longer freezes the stats page or an active video player; deletes are now batched into a single transaction.
+- Styled Subtitle Cue Parsing: Heavily typeset subtitles (karaoke, signs) no longer flood the subtitle sidebar with garbage; vector drawing commands are no longer shown as text, and duplicate/animation-burst cues now collapse into one.
+- X11 mpv Renderer: Fixes an mpv crash on the first fullscreen toggle for X11/XWayland users with `gpu-next` shaders (e.g. ArtCNN), which was caused by X11 mode forcing the legacy OpenGL renderer.
+- X11 Overlay Display Scaling: Fixes the overlay appearing oversized and offset from mpv on X11/XWayland under fractional or mixed-monitor display scaling.
+
+<details>
+<summary>Internal changes</summary>
+
+**Internal**
+- Subtitle text is now decoded from ASS exactly once at ingest, so the renderer, timing tracker, and tokenizer all share one decoded value instead of each re-deriving it.
+- Added per-stage debug timings (`scanMs`, `mecabMs`, `frequencyMs`, `annotateMs`) to the subtitle tokenization pipeline log.
+
+</details>
+
+## v0.19.2 (2026-08-04)
+
+**Changed**
+- Subsync: The sync modal now lets you choose both the reference subtitle (correct timing) and the out-of-sync subtitle to retime, for both alass and ffsubsync. alass can also use the loaded video's audio as a reference for local files. Retiming the secondary track now reloads the result into the secondary slot instead of overwriting the primary subtitle.
+
+**Fixed**
+- Streaming Subtitle Tokenization: Jellyfin streams now seed subtitle tokenization directly from the downloaded subtitle file instead of relying on an mpv event that could be missed, and prefetching now runs to the end of the file and clears between episodes. The tokenization cache was raised from 256 to 2500 lines, and parsed cues are no longer lost when the active subtitle track briefly can't be resolved (e.g. switching to an embedded track). Together these prevent episodes from falling back to slow, line-by-line tokenization during playback.
+- Overlay: Subtitle lines now appear immediately at their cue time even on a tokenization cache miss, upgrading in place once tokens and annotations are ready, instead of waiting on a line still being processed. A failed tokenization is no longer cached as plain text, so repeated lines get another chance at annotations.
+- Background Logging: Background startup now respects the configured logging level when no explicit log level is passed.
+
+<details>
+<summary>Internal changes</summary>
+
+**Internal**
+- Patched three high-severity dependency advisories (`undici`, `brace-expansion`, `fast-uri`).
+
+</details>
+
+## v0.19.1 (2026-08-01)
+
+**Added**
+- Word Card Type: Adds a setting (Settings > Mining/Anki > Kiku/Lapis Features > "Word Card Type") to choose which card-type flag SubMiner marks on Kiku/Lapis word cards — `word-and-sentence` (default), `click`, `sentence`, `audio`, or `none`. Click cards (`IsClickCard`) can now be flagged, and setting any card-type flag clears the others so a note can't claim two types at once.
+
+**Fixed**
+- Yomitan Popup: Fixes the macOS Yomitan popup going inert after mining a card — clicks outside the popup no longer pass through to mpv, and scrolling over the popup scrolls its definitions instead of seeking playback.
+- YouTube Playlist Links: Fixes opening a video from a playlist URL (e.g. a Watch Later link with `list=`/`index=`) timing out while probing subtitles, metadata, or the playback URL.
+
 ## v0.19.0 (2026-07-29)
 
 **Added**

docs-site/configuration.md

@@ -398,30 +398,30 @@ See `config.example.jsonc` for detailed configuration options.
 }
 ```
 
-| Option                             | Values   | Description                                                                                                                  |
+| Option                             | Values   | Description                                                                                                                                                               |
-| ---------------------------------- | -------- | ---------------------------------------------------------------------------------------------------------------------------- |
+| ---------------------------------- | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `primaryDefaultMode`               | string   | Default primary subtitle bar visibility mode: `"hidden"`, `"visible"`, or `"hover"` (default: `"visible"`)                   |
+| `primaryDefaultMode`               | string   | Default primary subtitle bar visibility mode: `"hidden"`, `"visible"`, or `"hover"` (default: `"visible"`)                                                                |
-| `subtitleStyle.css`                | object   | CSS declaration object applied to primary subtitles after normal style defaults. Use CSS property names such as `font-size`. |
+| `subtitleStyle.css`                | object   | CSS declaration object applied to primary subtitles after normal style defaults. Use CSS property names such as `font-size`.                                              |
-| `secondary.css`                    | object   | CSS declaration object applied to secondary subtitles after normal secondary style defaults.                                 |
+| `secondary.css`                    | object   | CSS declaration object applied to secondary subtitles after normal secondary style defaults.                                                                              |
-| `enableJlpt`                       | boolean  | Enable JLPT level underline styling (`false` by default)                                                                     |
+| `enableJlpt`                       | boolean  | Enable JLPT level underline styling (`false` by default)                                                                                                                  |
-| `preserveLineBreaks`               | boolean  | Preserve line breaks in visible overlay subtitle rendering (`false` by default). Enable to mirror mpv line layout.           |
+| `preserveLineBreaks`               | boolean  | Preserve line breaks in visible overlay subtitle rendering (`false` by default). Enable to mirror mpv line layout.                                                        |
-| `autoPauseVideoOnHover`            | boolean  | Pause playback while mouse hovers subtitle text, then resume on leave (`true` by default).                                   |
+| `autoPauseVideoOnHover`            | boolean  | Pause playback while mouse hovers subtitle text, then resume on leave (`true` by default).                                                                                |
-| `autoPauseVideoOnYomitanPopup`     | boolean  | Pause playback while the Yomitan popup is open, then resume when the popup closes (`true` by default).                       |
+| `autoPauseVideoOnYomitanPopup`     | boolean  | Pause playback while the Yomitan popup is open, then resume when the popup closes (`true` by default).                                                                    |
-| `primaryVisibleOnYomitanPopup`     | boolean  | Keep hover-mode primary subtitles visible while the Yomitan popup is open (`true` by default).                               |
+| `primaryVisibleOnYomitanPopup`     | boolean  | Keep hover-mode primary subtitles visible while the Yomitan popup is open (`true` by default).                                                                            |
-| `nameMatchEnabled`                 | boolean  | Enable character dictionary sync and subtitle token coloring for character-name matches (`false` by default)                 |
+| `nameMatchEnabled`                 | boolean  | Enable character dictionary sync and subtitle token coloring for character-name matches (`false` by default)                                                              |
-| `nameMatchImagesEnabled`           | boolean  | Show small cached AniList character portraits beside matched character-name tokens (`false` by default)                      |
+| `nameMatchImagesEnabled`           | boolean  | Show small cached AniList character portraits beside matched character-name tokens (`false` by default)                                                                   |
-| `nameMatchColor`                   | string   | Hex color used for subtitle tokens matched from the SubMiner character dictionary (default: `#f5bde6`)                       |
+| `nameMatchColor`                   | string   | Hex color used for subtitle tokens matched from the SubMiner character dictionary (default: `#f5bde6`)                                                                    |
-| `knownWordColor`                   | string   | Hex color used for known-word subtitle highlights (default: `#a6da95`)                                                       |
+| `knownWordColor`                   | string   | Hex color used for known-word subtitle highlights (default: `#a6da95`)                                                                                                    |
 | `knownWordMaturityColors`          | object   | Per-tier known-word colors used when `ankiConnect.knownWords.maturityEnabled` is on: `new` (`#ee99a0`), `learning` (`#b7bdf8`), `young` (`#91d7e3`), `mature` (`#a6da95`) |
-| `nPlusOneColor`                    | string   | Hex color used for the single N+1 target subtitle highlight (default: `#c6a0f6`)                                             |
+| `nPlusOneColor`                    | string   | Hex color used for the single N+1 target subtitle highlight (default: `#c6a0f6`)                                                                                          |
-| `frequencyDictionary.enabled`      | boolean  | Enable frequency highlighting from dictionary lookups (`false` by default)                                                   |
+| `frequencyDictionary.enabled`      | boolean  | Enable frequency highlighting from dictionary lookups (`false` by default)                                                                                                |
-| `frequencyDictionary.sourcePath`   | string   | Path to a local frequency dictionary root. Leave empty or omit to use installed/default frequency-dictionary search paths.   |
+| `frequencyDictionary.sourcePath`   | string   | Path to a local frequency dictionary root. Leave empty or omit to use installed/default frequency-dictionary search paths.                                                |
-| `frequencyDictionary.topX`         | number   | Only color tokens whose frequency rank is `<= topX` (`10000` by default)                                                     |
+| `frequencyDictionary.topX`         | number   | Only color tokens whose frequency rank is `<= topX` (`10000` by default)                                                                                                  |
-| `frequencyDictionary.mode`         | string   | `"single"` or `"banded"` (`"single"` by default)                                                                             |
+| `frequencyDictionary.mode`         | string   | `"single"` or `"banded"` (`"single"` by default)                                                                                                                          |
-| `frequencyDictionary.matchMode`    | string   | `"headword"` or `"surface"` (`"headword"` by default)                                                                        |
+| `frequencyDictionary.matchMode`    | string   | `"headword"` or `"surface"` (`"headword"` by default)                                                                                                                     |
-| `frequencyDictionary.singleColor`  | string   | Color used for all highlighted tokens in single mode                                                                         |
+| `frequencyDictionary.singleColor`  | string   | Color used for all highlighted tokens in single mode                                                                                                                      |
-| `frequencyDictionary.bandedColors` | string[] | Array of five hex colors used for ranked bands in banded mode                                                                |
+| `frequencyDictionary.bandedColors` | string[] | Array of five hex colors used for ranked bands in banded mode                                                                                                             |
-| `jlptColors`                       | object   | JLPT level underline colors object (`N1`..`N5`)                                                                              |
+| `jlptColors`                       | object   | JLPT level underline colors object (`N1`..`N5`)                                                                                                                           |
 
 Subtitle CSS custom properties:
 
@@ -555,11 +555,11 @@ Secondary subtitles do **not** auto-load by default. To turn them on for local a
 }
 ```
 
-| Option                  | Values                             | Description                                                                                                                                                             |
+| Option                  | Values                             | Description                                                                                                                                   |
-| ----------------------- | ---------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| ----------------------- | ---------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
-| `secondarySubLanguages` | string[]                           | Language codes to auto-load (e.g., `["eng", "en"]`); non-Signs/Songs tracks are preferred when several tracks match. Default is empty (`[]`).                            |
+| `secondarySubLanguages` | string[]                           | Language codes to auto-load (e.g., `["eng", "en"]`); non-Signs/Songs tracks are preferred when several tracks match. Default is empty (`[]`). |
-| `autoLoadSecondarySub`  | `true`, `false`                    | Auto-detect and load a matching secondary subtitle track for local/Jellyfin sidecar files (default: `false`)                                                             |
+| `autoLoadSecondarySub`  | `true`, `false`                    | Auto-detect and load a matching secondary subtitle track for local/Jellyfin sidecar files (default: `false`)                                  |
-| `defaultMode`           | `"hidden"`, `"visible"`, `"hover"` | Initial display mode (default: `"hover"`)                                                                                                                               |
+| `defaultMode`           | `"hidden"`, `"visible"`, `"hover"` | Initial display mode (default: `"hover"`)                                                                                                     |
 
 These two settings apply to local and Jellyfin playback only. YouTube secondary selection is fixed to English and ignores them; see [YouTube Integration](/youtube-integration#secondary-subtitle-languages). `defaultMode` still controls how the loaded secondary bar is displayed in every case.
 
@@ -1043,7 +1043,7 @@ This example is intentionally compact. The option table below documents availabl
 | `ankiConnect.knownWords.refreshMinutes`           | number                                      | Minutes between known-word cache refreshes (default: `1440`)                                                                                                                                                                    |
 | `ankiConnect.knownWords.decks`                    | object                                      | Deck→fields mapping used for known-word cache query scope (e.g. `{ "Kaishi 1.5k": ["Word"] }`).                                                                                                                                 |
 | `ankiConnect.knownWords.maturityEnabled`          | `true`, `false`                             | Color known words by Anki card maturity (new/learning/young/mature) instead of one color. Requires `knownWords.highlightEnabled` (default: `false`). Tier colors come from `subtitleStyle.knownWordMaturityColors`.             |
-| `ankiConnect.knownWords.matureThresholdDays`      | number                                      | Card interval in days at which a known word counts as mature (default: `21`, matching Anki's own convention)                                                                                                                   |
+| `ankiConnect.knownWords.matureThresholdDays`      | number                                      | Card interval in days at which a known word counts as mature (default: `21`, matching Anki's own convention)                                                                                                                    |
 | `ankiConnect.nPlusOne.enabled`                    | `true`, `false`                             | Enable N+1 subtitle highlighting (highlights the one unknown word in a sentence). Independent from `knownWords.highlightEnabled`. Requires known-word cache data (default: `false`).                                            |
 | `ankiConnect.nPlusOne.minSentenceWords`           | number                                      | Minimum number of words required in a sentence before single unknown-word N+1 highlighting can trigger (default: `3`).                                                                                                          |
 | `behavior.notificationType`                       | `"overlay"`, `"system"`, `"both"`, `"none"` | Notification type on card update (default: `"overlay"`). `"both"` means overlay + system. `osd` and `osd-system` are legacy config-file-only values; use `"osd-system"` to keep the old OSD + system behavior.                  |
@@ -1069,6 +1069,9 @@ SubMiner is intentionally built for [Kiku](https://kiku.youyoumu.my.id/) and [La
     "enabled": true,
     "fieldGrouping": "manual",
     "deleteDuplicateInAuto": true
+  },
+  "lapisKiku": {
+    "wordCardKind": "word-and-sentence"
   }
 }
 ```
@@ -1077,6 +1080,21 @@ SubMiner is intentionally built for [Kiku](https://kiku.youyoumu.my.id/) and [La
 - Enable `isKiku` to turn on duplicate merge behavior for mined Word/Expression hits.
 - When both are enabled, Kiku behavior is applied for grouping while sentence-card model settings are still read from `isLapis`.
 - `isKiku.fieldGrouping` supports `disabled`, `auto`, and `manual` merge modes; see [Field Grouping Modes](#field-grouping-modes).
+- `lapisKiku.wordCardKind` picks the card-type flag set on word cards; see [Word Card Type](#word-card-type). It is read only while `isLapis` or `isKiku` is enabled.
+
+### Word Card Type
+
+When SubMiner fills the sentence on a mined word card - from Yomitan auto-enrichment, a manual clipboard update, or stats-dashboard word mining - it marks which card that note should generate. `ankiConnect.lapisKiku.wordCardKind` chooses the flag:
+
+| Value                         | Flag set                |
+| ----------------------------- | ----------------------- |
+| `word-and-sentence` (default) | `IsWordAndSentenceCard` |
+| `click`                       | `IsClickCard`           |
+| `sentence`                    | `IsSentenceCard`        |
+| `audio`                       | `IsAudioCard`           |
+| `none`                        | none; flags left as-is  |
+
+The other card-type flags are cleared so a note never claims two card types at once. Notes are skipped when the note type has no field for the chosen flag, and when the note was already mined as a sentence or audio card. Cards created by Mine Sentence and Mine Audio keep their own flag regardless of this setting.
 
 ### N+1 Word Highlighting
 
@@ -1167,10 +1185,10 @@ TsukiHime subtitle search works out of the box and needs no account or API key.
 }
 ```
 
-| Option                       | Values       | Description                                                                                          |
+| Option                       | Values       | Description                                                                                           |
-| ---------------------------- | ------------ | ---------------------------------------------------------------------------------------------------- |
+| ---------------------------- | ------------ | ----------------------------------------------------------------------------------------------------- |
 | `tsukihime.apiBaseUrl`       | string (URL) | Base URL of the TsukiHime API (default: `https://api.tsukihime.org/v1`). Only change it for a mirror. |
-| `tsukihime.maxSearchResults` | number       | Maximum releases returned per search (default: `10`; the API caps this at 100)                       |
+| `tsukihime.maxSearchResults` | number       | Maximum releases returned per search (default: `10`; the API caps this at 100)                        |
 
 The keyboard shortcut lives under `shortcuts.openTsukihime` (default `Ctrl+Shift+T`; set to `null` to disable). The older `animetosho` section and `shortcuts.openAnimetosho` are still accepted as deprecated aliases, with the current names taking precedence when both are set.
 
@@ -1178,9 +1196,9 @@ See [TsukiHime Integration](/tsukihime-integration) for the modal workflow, lang
 
 ### Subtitle Sync
 
-Sync the active subtitle track from the overlay picker using `alass` or `ffsubsync`. Both are **optional external tools** that must be installed separately and available on your `PATH` (or configured via the path options below).
+Sync a subtitle track from the overlay picker using `alass` or `ffsubsync`. The picker lets you choose which track gets retimed (the active primary track by default) and, for alass, which reference it is aligned against (the secondary subtitle track by default). Both are **optional external tools** that must be installed separately and available on your `PATH` (or configured via the path options below).
 
-- [`alass`](https://github.com/kaegi/alass) - fast, audio-independent sync using a secondary subtitle as reference
+- [`alass`](https://github.com/kaegi/alass) - fast, audio-independent sync using another subtitle as reference; it can also take the local video file as reference (alass extracts the audio itself)
 - [`ffsubsync`](https://github.com/smacke/ffsubsync) - audio-based sync using the video file as reference
 
 ```json
@@ -1228,17 +1246,17 @@ AniList integration is opt-in and disabled by default. Enable it to allow SubMin
 }
 ```
 
-| Option                                                         | Values              | Description                                                                                                   |
+| Option                                                         | Values                  | Description                                                                                                   |
-| -------------------------------------------------------------- | ------------------- | ------------------------------------------------------------------------------------------------------------- |
+| -------------------------------------------------------------- | ----------------------- | ------------------------------------------------------------------------------------------------------------- |
-| `anilist.enabled`                                              | `true`, `false`     | Enable AniList post-watch progress updates (default: `false`)                                                 |
+| `anilist.enabled`                                              | `true`, `false`         | Enable AniList post-watch progress updates (default: `false`)                                                 |
-| `accessToken`                                                  | string              | Optional explicit AniList access token override (default: empty string)                                       |
+| `accessToken`                                                  | string                  | Optional explicit AniList access token override (default: empty string)                                       |
-| `characterDictionary.maxLoaded`                                | number              | Maximum number of most-recently-used AniList media snapshots included in the merged dictionary (default: `3`) |
+| `characterDictionary.maxLoaded`                                | number                  | Maximum number of most-recently-used AniList media snapshots included in the merged dictionary (default: `3`) |
-| `characterDictionary.refreshTtlHours`                          | number              | Hours before a cached media snapshot is refreshed (default: `168`, clamped to 1–8760)                         |
+| `characterDictionary.refreshTtlHours`                          | number                  | Hours before a cached media snapshot is refreshed (default: `168`, clamped to 1–8760)                         |
-| `characterDictionary.evictionPolicy`                           | `"delete"`, `"disable"` | What happens to snapshots evicted beyond `maxLoaded` (default: `"delete"`)                                |
+| `characterDictionary.evictionPolicy`                           | `"delete"`, `"disable"` | What happens to snapshots evicted beyond `maxLoaded` (default: `"delete"`)                                    |
-| `characterDictionary.collapsibleSections.description`          | `true`, `false`     | Open the Description section by default in generated dictionary entries                                       |
+| `characterDictionary.collapsibleSections.description`          | `true`, `false`         | Open the Description section by default in generated dictionary entries                                       |
-| `characterDictionary.collapsibleSections.characterInformation` | `true`, `false`     | Open the Character Information section by default in generated dictionary entries                             |
+| `characterDictionary.collapsibleSections.characterInformation` | `true`, `false`         | Open the Character Information section by default in generated dictionary entries                             |
-| `characterDictionary.collapsibleSections.voicedBy`             | `true`, `false`     | Open the Voiced by section by default in generated dictionary entries                                         |
+| `characterDictionary.collapsibleSections.voicedBy`             | `true`, `false`         | Open the Voiced by section by default in generated dictionary entries                                         |
-| `characterDictionary.profileScope`                             | `"all"`, `"active"` | Apply dictionary settings updates to all Yomitan profiles or only active profile                              |
+| `characterDictionary.profileScope`                             | `"all"`, `"active"`     | Apply dictionary settings updates to all Yomitan profiles or only active profile                              |
 
 When `enabled` is `true` and `accessToken` is empty, SubMiner opens an AniList setup helper window. Keep `enabled` as `false` to disable all AniList setup/update behavior.
 
@@ -1539,18 +1557,18 @@ Configure the mpv executable, profile, and window state for SubMiner-managed mpv
 }
 ```
 
-| Option                   | Values                                                                      | Description                                                                                                                                 |
+| Option                   | Values                                                                      | Description                                                                                                                                                                         |
-| ------------------------ | --------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------- |
+| ------------------------ | --------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `executablePath`         | string                                                                      | Absolute path to `mpv.exe` for Windows launch flows. Leave empty to auto-discover from `SUBMINER_MPV_PATH` or `PATH` (default `""`)         |
+| `executablePath`         | string                                                                      | Absolute path to `mpv.exe` for Windows launch flows. Leave empty to auto-discover from `SUBMINER_MPV_PATH` or `PATH` (default `""`)                                                 |
-| `profile`                | string                                                                      | mpv profile name passed as `--profile=<name>`. Leave empty to pass no profile (default `""`)                                                |
+| `profile`                | string                                                                      | mpv profile name passed as `--profile=<name>`. Leave empty to pass no profile (default `""`)                                                                                        |
-| `launchMode`             | `"normal"` \| `"maximized"` \| `"fullscreen"`                               | Window state when SubMiner spawns mpv (default `"normal"`)                                                                                  |
+| `launchMode`             | `"normal"` \| `"maximized"` \| `"fullscreen"`                               | Window state when SubMiner spawns mpv (default `"normal"`)                                                                                                                          |
 | `socketPath`             | string                                                                      | mpv IPC socket path used by SubMiner-managed playback and the bundled mpv plugin (platform-dependent default: `/tmp/subminer-socket`, or `\\\\.\\pipe\\subminer-socket` on Windows) |
-| `backend`                | `"auto"` \| `"hyprland"` \| `"sway"` \| `"x11"` \| `"macos"` \| `"windows"` | Window tracking backend passed to the bundled mpv plugin. Auto detects the current platform (default: `"auto"`)                             |
+| `backend`                | `"auto"` \| `"hyprland"` \| `"sway"` \| `"x11"` \| `"macos"` \| `"windows"` | Window tracking backend passed to the bundled mpv plugin. Auto detects the current platform (default: `"auto"`)                                                                     |
-| `autoStartSubMiner`      | `true`, `false`                                                             | Start SubMiner in the background when SubMiner-managed mpv loads a file (default: `true`)                                                   |
+| `autoStartSubMiner`      | `true`, `false`                                                             | Start SubMiner in the background when SubMiner-managed mpv loads a file (default: `true`)                                                                                           |
-| `pauseUntilOverlayReady` | `true`, `false`                                                             | Pause mpv on visible-overlay auto-start until SubMiner signals subtitle tokenization readiness, with a 30-second fallback (default: `true`) |
+| `pauseUntilOverlayReady` | `true`, `false`                                                             | Pause mpv on visible-overlay auto-start until SubMiner signals subtitle tokenization readiness, with a 30-second fallback (default: `true`)                                         |
-| `subminerBinaryPath`     | string                                                                      | SubMiner app binary path passed to the bundled mpv plugin. Leave empty to use the launcher-detected app path (default: `""`)                |
+| `subminerBinaryPath`     | string                                                                      | SubMiner app binary path passed to the bundled mpv plugin. Leave empty to use the launcher-detected app path (default: `""`)                                                        |
-| `aniskipEnabled`         | `true`, `false`                                                             | Enable AniSkip intro detection, chapter markers, and the skip-intro key (default: `true`)                                                   |
+| `aniskipEnabled`         | `true`, `false`                                                             | Enable AniSkip intro detection, chapter markers, and the skip-intro key (default: `true`)                                                                                           |
-| `aniskipButtonKey`       | string                                                                      | mpv key used to skip the detected intro while the skip prompt is visible (default: `"TAB"`)                                                 |
+| `aniskipButtonKey`       | string                                                                      | mpv key used to skip the detected intro while the skip prompt is visible (default: `"TAB"`)                                                                                         |
 
 If `mpv.profile` is configured and the launcher also receives `--profile`, SubMiner passes both as a comma-separated mpv profile list.
 

docs-site/immersion-tracking.md

@@ -57,6 +57,13 @@ Jellyfin stream URLs are normalized to stable item links before stats titles are
 
 When YouTube channel metadata is available, the Library tab groups videos by creator/channel and treats each tracked video as an episode-like entry inside that channel section.
 
+A library entry is identified by its parsed title plus any detected season, so the same show can end up on several cards when releases disagree about the title or omit the season tag. Two fixes are available:
+
+- **Merge duplicates.** Hit **Select** above the grid, tick the cards that are the same show, and choose **Merge Selected**. Pick which entry to keep in the dialog; every episode moves onto it and the other cards are removed. Nothing is deleted, so sessions, mined cards and watch time all carry over. SubMiner remembers the merged title variants, so future episodes parsed with one of those names join the kept entry instead of recreating a duplicate card.
+- **Move a single episode.** Hover an episode row in a title's episode list and use the **→** button to reassign it to another library entry. The correction is remembered, so later filename parsing or Jellyfin metadata cannot move that episode back. For local files, later episodes in the same directory inherit the correction when their detected seasons are compatible and every manual correction there points to the same entry. Conflicting seasons or manual destinations are left for review. If the move empties the old entry, that card is removed and you are returned to the grid.
+
+Once cover art resolves a series to an AniList entry, cards with compatible seasons are folded together automatically only when the searched title exactly matches an AniList title or synonym. A fuzzy result that points at an AniList entry already used by another card appears as a **Possible duplicate** review above the Library grid instead. Choose **Review merge** to compare the cards and pick which one to keep, or **Not duplicates** to dismiss that suggestion permanently. Entries with conflicting explicit season numbers are left alone rather than merged or suggested.
+
 Open a title and use **Delete Entry** in its header to remove a mistakenly tracked show outright. This deletes every episode of that title along with their sessions, subtitle lines, rollups and cover art, drops the words and kanji that were only seen there, and removes the card from the Library grid. Individual episodes and sessions can still be deleted on their own from the episode list and session rows. Entry deletion is refused while that title is the one currently playing.
 
 ![Stats Library](/screenshots/stats-library.png)

docs-site/mining-workflow.md

@@ -161,10 +161,13 @@ If your subtitle file is out of sync with the audio, SubMiner can resynchronize
 
 1. Open the subsync modal from the overlay.
 2. Select the sync engine (alass or ffsubsync).
-3. For alass, select a reference subtitle track from the video.
+3. For alass, pick the **reference** - the subtitle with correct timing. This defaults to the secondary subtitle track. The loaded video file can also be used as the reference (alass extracts the audio itself), but it is never the default.
-4. SubMiner runs the sync and reloads the corrected subtitle.
+4. Pick the **out-of-sync subtitle** - the track that gets retimed. This defaults to the active primary subtitle track and applies to both engines.
+5. SubMiner runs the sync and reloads the corrected subtitle into the slot the out-of-sync track came from: retiming the secondary track keeps it secondary and leaves the primary track selected.
 
-For remote streams, including Jellyfin playback, the modal only offers alass. Jellyfin subtitle URLs are cached as temporary subtitle files so alass can read them, but the video stream is not downloaded. ffsubsync needs direct access to the local media file and is unavailable for stream URLs.
+The reference and the out-of-sync subtitle must be different tracks; the reference list hides whichever track is selected as the target.
+
+For remote streams, including Jellyfin playback, the modal only offers alass with a subtitle reference. Jellyfin subtitle URLs are cached as temporary subtitle files so alass can read them, but the video stream is not downloaded. ffsubsync and the video-file reference need direct access to the local media file and are unavailable for stream URLs.
 
 When you mine a sentence card from the stats dashboard, SubMiner can also use `alass` automatically to align a local English sidecar against the matching local Japanese sidecar before filling the card translation field. The source subtitle files are not modified; SubMiner writes a temporary retimed copy and reuses it while the stats server is running.
 

docs-site/public/config.example.jsonc

@@ -523,7 +523,7 @@
   // ==========================================
   // AnkiConnect Integration
   // Automatic Anki updates and media generation options.
-  // Hot-reload: ankiConnect.ai.enabled, media.normalizeAudio/mirrorMpvVolume, knownWords, nPlusOne, fields.word/audio/image/sentence/miscInfo, behavior.autoUpdateNewCards, isLapis.sentenceCardModel, and isKiku.fieldGrouping update live while SubMiner is running.
+  // Hot-reload: ankiConnect.ai.enabled, media.normalizeAudio/mirrorMpvVolume, knownWords, nPlusOne, fields.word/audio/image/sentence/miscInfo, behavior.autoUpdateNewCards, isLapis.sentenceCardModel, isKiku.fieldGrouping, and lapisKiku.wordCardKind update live while SubMiner is running.
   // Shared AI provider transport settings are read from top-level ai and typically require restart.
   // Most other AnkiConnect settings still require restart.
   // ==========================================
@@ -605,7 +605,10 @@
       "enabled": false, // Enable Kiku-specific mining behaviors (duplicate handling, field grouping). Values: true | false
       "fieldGrouping": "disabled", // Kiku duplicate-card field grouping mode. Values: auto | manual | disabled
       "deleteDuplicateInAuto": true // When Kiku field grouping is "auto", delete the duplicate source card after grouping completes. Values: true | false
-    } // Is kiku setting.
+    }, // Is kiku setting.
+    "lapisKiku": {
+      "wordCardKind": "word-and-sentence" // Card-type flag SubMiner marks on Kiku/Lapis word cards. Only one flag is set at a time; the others are cleared. Requires isKiku.enabled or isLapis.enabled. Values: word-and-sentence | click | sentence | audio | none
+    } // Lapis kiku setting.
   }, // Automatic Anki updates and media generation options.
 
   // ==========================================

docs-site/troubleshooting.md

@@ -227,7 +227,7 @@ Install ffsubsync or configure the path:
 
 If subtitle sync fails (the error message is prefixed with the engine name):
 
-- Ensure the reference subtitle track exists in the video (alass requires a source track).
+- Ensure a reference is selected (alass needs either a second subtitle track or the local video file, and it cannot be the same track that is being retimed).
 - Check that `ffmpeg` is available (used to extract the internal subtitle track).
 - Try running the sync tool manually to see detailed error output.
 - ffsubsync requires local files and cannot handle remote media streams (e.g., streaming URLs).
@@ -405,8 +405,9 @@ On any Wayland session that is not Hyprland or Sway (KDE Plasma, GNOME, and othe
 
 SubMiner handles this automatically:
 
-- It launches its own window under XWayland (it sets `--ozone-platform-hint=x11`).
+- It launches its own window under XWayland (it sets `--ozone-platform=x11`).
-- Every mpv it launches (via the `subminer` launcher, Jellyfin, or YouTube) is pinned to XWayland too - Wayland environment hints are stripped and an X11 GPU context (`--gpu-context=x11egl,x11`) is applied.
+- Every mpv it launches (via the `subminer` launcher, Jellyfin, or YouTube) is pinned to XWayland too - Wayland environment hints are stripped and an X11 GPU context (`--gpu-context=x11vk,x11egl,x11`) is applied. Only the window context is overridden; your `vo`/`gpu-api` and user shaders are left alone.
+- Fractional and mixed-monitor display scaling is handled per screen when SubMiner maps XWayland mpv coordinates to the overlay.
 - While mpv is windowed, the overlay is a managed X11 window owned by the tracked mpv window (`WM_TRANSIENT_FOR`), so it stays above mpv while other foreground X11/Xwayland apps can still cover both windows.
 - While tracked mpv is fullscreen, SubMiner swaps the visible overlay to a focusable-false X11 override-redirect window. That path can stay above the active fullscreen mpv window without requiring a KDE/KWin-specific rule, and SubMiner hides/releases it when mpv is no longer the active X11/Xwayland window.
 - The visible overlay is shown inactive on Linux, so normal hover should not steal keyboard focus from mpv.
@@ -420,7 +421,7 @@ Requirements: `xdotool`, `xprop`, and `xwininfo` must be installed. SubMiner use
 This almost always means mpv came up as a **native Wayland** window that the XWayland overlay cannot cover. It happens when mpv is launched **manually** (your own command), because SubMiner can only force XWayland on the mpv processes it launches itself. Fix it one of these ways:
 
 - Launch playback through SubMiner (the `subminer` launcher or the tray), which forces XWayland for you, or
-- Force XWayland in your own mpv invocation, e.g. `mpv --gpu-context=x11egl …`, or launch with `WAYLAND_DISPLAY= mpv …`, or set `gpu-context=x11egl` in your `mpv.conf`.
+- Force XWayland in your own mpv invocation, e.g. `mpv --gpu-context=x11vk,x11egl,x11 …`, or launch with `WAYLAND_DISPLAY= mpv …`, or set `gpu-context=x11vk` (Vulkan) / `gpu-context=x11egl` (OpenGL) in your `mpv.conf`.
 
 To confirm mpv is on XWayland, `xdotool search --class mpv` should return a window id (a native Wayland mpv returns nothing).
 

docs-site/usage.md

@@ -145,11 +145,13 @@ The tray menu includes `Export Logs`, which creates the same sanitized local-dat
 
 Once Jellyfin is configured, the tray menu includes `Jellyfin Discovery` for starting or stopping cast discovery in the current app session without changing config.
 
+The tray menu also includes `View Changelog`, which opens the in-app changelog modal. It fetches the changelog from the newest published release, so you see release notes for versions newer than the one you run; if the download fails it falls back to the changelog bundled with your install and says so. Versions in the current `0.x` line are expanded by default and older lines are folded, matching this site's [Changelog](/changelog). A badge marks the version you have installed, and newer versions are tagged `New`. The same modal opens from the `What's New` button on the update-available overlay notification.
+
 ### Logging and App Mode
 
 - `--log-level` controls logger verbosity.
 - `--dev` and `--debug` are app/dev-mode switches; they are not log-level aliases.
-- `--background` defaults to quieter logging (`warn`) unless `--log-level` is set.
+- `--background` starts at the default quieter logging level (`warn`), then follows `logging.level` after config loads. An explicit `--log-level` remains the override.
 - `--background` launched from a terminal detaches and returns the prompt; stop it with tray Quit or `SubMiner.AppImage --stop` (`SubMiner.exe --stop` on Windows).
 - Linux desktop launcher starts SubMiner with `--background` by default (via electron-builder `linux.executableArgs`).
 - On Hyprland and other Wayland compositors, the tray icon appears only when your panel provides a StatusNotifier/AppIndicator tray host.
@@ -368,6 +370,8 @@ Press `V` to cycle the primary SubMiner subtitle bar through hidden → visible
 
 `Ctrl/Cmd+/` opens the session help modal with the current overlay and mpv keybindings. The same help view is also available through the `y-h` chord in mpv.
 
+The changelog modal (tray > `View Changelog`) works the same way: it renders over mpv when a video is playing and in its own window otherwise. Use `J`/`K` or the arrow keys to move between versions, `Enter` to fold or unfold one, `R` to refetch, and `Esc` to close.
+
 Hovering over subtitle text pauses mpv by default; leaving resumes it. Yomitan popups also pause playback by default. Set `subtitleStyle.autoPauseVideoOnHover: false` or `subtitleStyle.autoPauseVideoOnYomitanPopup: false` to disable either behavior.
 
 ### Drag-and-Drop

docs-site/websocket-texthooker-api.md

@@ -64,7 +64,7 @@ Use the basic subtitle websocket when you only need the current subtitle line as
 - **Client auth:** none
 - **Reconnects:** client-managed
 
-When a client connects, SubMiner immediately sends the latest subtitle payload if one is available. After that, it pushes a new message each time the current subtitle changes.
+When a client connects, SubMiner immediately sends the latest subtitle payload if one is available. After that, it pushes a new message each time the current subtitle changes. Annotation-only upgrades do not repeat the same line on this basic stream.
 
 #### Message shape
 
@@ -96,6 +96,8 @@ Use the annotation websocket for custom clients that want the same structured to
 
 In practice, if you are building a new client, prefer `annotationWebsocket` unless you specifically need compatibility with an existing `websocket` consumer.
 
+On a tokenization cache miss, this stream first sends the cue as plain text with an empty `tokens` array, then sends the annotated replacement when tokenization finishes. Treat each message as the complete current state, replacing the previous payload.
+
 #### Message shape
 
 ```json

docs-site/youtube-integration.md

@@ -11,7 +11,7 @@ SubMiner auto-loads Japanese subtitles when you play a YouTube URL, giving you t
 
 When SubMiner detects a YouTube URL (or `ytsearch:` target), it pauses mpv at startup and runs a subtitle pipeline before resuming playback:
 
-1. **Probe** --- `yt-dlp --dump-single-json` extracts all available subtitle tracks (manual uploads and auto-generated captions) along with video metadata.
+1. **Probe** --- `yt-dlp --dump-single-json` extracts all available subtitle tracks (manual uploads and auto-generated captions) along with video metadata. Every yt-dlp call passes `--no-playlist`, so playlist links (for example a Watch Later URL with `list=`/`index=`) resolve to the single video instead of the whole playlist.
 2. **Discover** --- Each track is normalized into a `YoutubeTrackOption` with language code, kind (`manual` or `auto`), display label, and direct download URL.
 3. **Select** --- SubMiner picks the best primary track (Japanese, preferring manual over auto) and secondary track (English, preferring manual over auto).
 4. **Download** --- Selected tracks are fetched via direct URL when available, falling back to `yt-dlp --write-subs` / `--write-auto-subs`. YouTube TimedText XML formats (`srv1`/`srv2`/`srv3`) are converted to VTT on the fly. Auto-generated VTT captions are normalized to remove rolling-caption duplication.

docs/architecture/2026-03-15-renderer-performance-design.md

@@ -64,18 +64,23 @@ External subtitle files only (SRT, VTT, ASS). Embedded subtitle tracks are out o
 A cue parser extracts both timing and text content from subtitle files for prefetching.
 
 **Parsed cue structure:**
+
 ```typescript
 interface SubtitleCue {
-  startTime: number;  // seconds
+  startTime: number; // seconds
-  endTime: number;    // seconds
+  endTime: number; // seconds
-  text: string;       // raw subtitle text
+  text: string; // plain text, decoded from the source format
 }
 ```
 
 **Supported formats:**
+
 - SRT/VTT: Regex-based parsing of timing lines + text content between timing blocks.
-- ASS: Parse `[Events]` section, extract `Dialogue:` lines, split on the first 9 commas only (ASS v4+ has 10 fields; the last field is Text which can itself contain commas). Strip ASS override tags (`{\...}`) from the text before storing.
+- ASS: Parse `[Events]` section, extract `Dialogue:` lines, read the field order from the `Format:` row, and take everything after the Text field index as the text (Text can itself contain commas).
-  ASS text fields contain inline override tags like `{\b1}`, `{\an8}`, `{\fad(200,300)}`. The cue parser strips these during extraction so the tokenizer receives clean text.
+
+**ASS decoding.** The parser is where ASS text is decoded, once, via `assToPlainText()` in `src/core/services/ass-text.ts`. That decoder mirrors mpv's `ass_to_plaintext` so a cue read from a file reads identically to the same line arriving live on `sub-text`: `{...}` override blocks are markup, `\pN … \p0` vector drawing runs are dropped rather than shown as text, `\N`/`\n`/`\h` are the only escapes (`\{`, `\}` and `\\` are not), and an unclosed `{` is rendered verbatim. Every layer downstream — renderer, timing tracker, tokenizer, tokenization cache keys — receives plain text and uses `normalizePlainSubtitleText()` for whitespace only, so nothing decodes the same string twice and one authored line always maps to one cache key.
+
+**Duplicate collapsing.** Typeset scripts emit one `Dialogue:` event per animation frame, plus layered copies of the same line. The parser collapses identical text over an identical span unconditionally, and collapses contiguous same-text runs of at least three events when the run looks like an animation. For ASS that means shared style and actor plus authoring evidence: a temporal tag (`\t`, `\move`, `\k`/`\kf`/`\ko`/`\K`, or anything wrapped in `\t(...)`), an animated `Effect` column (`Karaoke`, `Banner`, `Scroll`), or override values that change across the run. Static tags shared by every event (`\pos`, an identical `\clip`) are not evidence. SRT/VTT carry no such metadata, so there collapsing needs at least five contiguous events all under 0.1s — the frame timing left behind by ASS-to-SRT conversion. The parser keeps this authoring metadata (style, actor, layer, `Effect`, parsed override commands, source order) private; `parseSubtitleCues()` returns only `SubtitleCue`.
 
 #### Prefetch Service Lifecycle
 
@@ -153,6 +158,7 @@ tokens (already have frequencyRank values from parser-level applyFrequencyRanks)
 ### Dependency Analysis
 
 All annotations either depend on MeCab POS data or benefit from running after it:
+
 - **Known word marking:** Needs base tokens (surface/headword). No POS dependency, but no reason to run separately.
 - **Frequency filtering:** Uses `pos1Exclusions` and `pos2Exclusions` to clear frequency ranks on excluded tokens (particles, noise). Depends on MeCab POS data.
 - **JLPT marking:** Uses `shouldIgnoreJlptForMecabPos1` to filter. Depends on MeCab POS data.
@@ -169,18 +175,14 @@ function annotateTokens(tokens, deps, options): MergedToken[] {
 
   // Single pass: known word + frequency filtering + JLPT computed together
   const annotated = tokens.map((token) => {
-    const isKnown = nPlusOneEnabled
+    const isKnown = nPlusOneEnabled ? token.isKnown || computeIsKnown(token, deps) : false;
-      ? token.isKnown || computeIsKnown(token, deps)
-      : false;
 
     // Filter frequency rank using POS exclusions (rank values already set at parser level)
     const frequencyRank = frequencyEnabled
       ? filterFrequencyRank(token, pos1Exclusions, pos2Exclusions)
       : undefined;
 
-    const jlptLevel = jlptEnabled
+    const jlptLevel = jlptEnabled ? computeJlptLevel(token, deps.getJlptLevel) : undefined;
-      ? computeJlptLevel(token, deps.getJlptLevel)
-      : undefined;
 
     return { ...token, isKnown, frequencyRank, jlptLevel };
   });
@@ -221,6 +223,7 @@ Replace `document.createElement('span')` calls in the renderer with `templateSpa
 ### Current Behavior
 
 In `renderWithTokens` (`subtitle-render.ts`), each render cycle:
+
 1. Clears DOM with `innerHTML = ''`
 2. Creates a `DocumentFragment`
 3. Calls `document.createElement('span')` for each token (~10-15 per subtitle)
@@ -256,27 +259,30 @@ Full recycling (collecting old nodes, clearing attributes, reusing them) require
 
 ## Combined Impact Summary
 
-| Scenario | Before | After | Improvement |
+| Scenario                          | Before     | After      | Improvement |
-|----------|--------|-------|-------------|
+| --------------------------------- | ---------- | ---------- | ----------- |
-| Normal playback (prefetch-warmed) | ~200-320ms | ~30-50ms | ~80-85% |
+| Normal playback (prefetch-warmed) | ~200-320ms | ~30-50ms   | ~80-85%     |
-| Cache hit (repeated subtitle) | ~72ms | ~55-65ms | ~10-20% |
+| Cache hit (repeated subtitle)     | ~72ms      | ~55-65ms   | ~10-20%     |
-| Cache miss (immediate seek) | ~200-320ms | ~150-260ms | ~20-25% |
+| Cache miss (immediate seek)       | ~200-320ms | ~150-260ms | ~20-25%     |
 
 ---
 
 ## Files Summary
 
 ### New Files
+
 - `src/core/services/subtitle-prefetch.ts`
 - `src/core/services/subtitle-cue-parser.ts`
 
 ### Modified Files
+
 - `src/core/services/subtitle-processing-controller.ts` (expose `preCacheTokenization`)
 - `src/core/services/tokenizer/annotation-stage.ts` (batched single-pass)
 - `src/renderer/subtitle-render.ts` (template cloneNode)
 - `src/main.ts` (wire up prefetch service)
 
 ### Test Files
+
 - New tests for subtitle cue parser (SRT, VTT, ASS formats)
 - New tests for subtitle prefetch service (priority window, seek, pause/resume)
 - Updated tests for annotation stage (same behavior, new implementation)

docs/architecture/domains.md

@@ -25,6 +25,8 @@ Read when: you need to find the owner module for a behavior or test surface
 - Anki workflow: `src/anki-integration/`, `src/core/services/anki-jimaku*.ts`
 - Immersion tracking: `src/core/services/immersion-tracker/`
   Includes stats storage/query schema such as `imm_videos`, `imm_media_art`, and `imm_youtube_videos` for per-video and YouTube-specific library metadata.
+  Library-entry identity aliases and merge recommendations are persisted alongside this schema; the stats HTTP and SPA layers only expose and present those domain decisions.
+  `delete-maintenance-scheduler.ts` coalesces and serializes stats deletes; expensive deletion and summary rebuilds run in `delete-maintenance-worker-thread.ts` while the tracker queues playback writes. Each batch uses one transaction, lexical update, rollup refresh, and lifetime rebuild.
 - AniList tracking + character dictionary: `src/core/services/anilist/`, `src/main/runtime/composers/anilist-*`, `src/main/character-dictionary-runtime.ts`, `src/main/character-dictionary-runtime/`
 - Jellyfin integration: `src/core/services/jellyfin*.ts`, `src/main/runtime/composers/jellyfin-*`
 - Window trackers: `src/window-trackers/`

docs/architecture/subtitle-overlay-priming.md

@@ -3,7 +3,7 @@
 # Subtitle Overlay Priming
 
 Status: active
-Last verified: 2026-06-14
+Last verified: 2026-08-04
 Owner: Kyle Yasuda
 Read when: debugging subtitle state or blank Linux/X11 overlay windows when the visible overlay is shown or recreated
 
@@ -47,18 +47,43 @@ subtitles do not draw.
    `emitSubtitle(payload)` and `refreshCurrentSubtitle(text)`, then prime secondary subtitles.
 6. Tokenization cache hit: call `consumeCachedSubtitle(text)`, `onSubtitleChange(text)`, and
    `emitSubtitle(cachedPayload)`, then prime secondary subtitles.
-7. Cache miss: call `refreshCurrentSubtitle(text)` and let normal tokenization emit the final
+7. Cache miss: call `refreshCurrentSubtitle(text)`. Normal processing emits a plain payload
-   payload.
+   synchronously, then replaces it with the tokenized payload when ready.
 
-In `src/main.ts`, both `onSubtitleChange` and `refreshCurrentSubtitle` pause
+Both `onSubtitleChange` and `refreshCurrentSubtitle` pause `subtitlePrefetchService` and then call
-`subtitlePrefetchService`, notify it with `onSeek(lastObservedTimePos)`, and then call the matching
+the matching `subtitleProcessingController` method, giving the visible overlay priority over
-`subtitleProcessingController` method. This gives the visible overlay priority over background
+background prefetch work. Prefetch is not re-centered here: restarting the run per line
-prefetch work and re-centers prefetch around the live playback time.
+(`onSeek`) discarded the in-flight tokenization every time the subtitle changed, so only real
+seeks restart it (see `onTimePosUpdate` in `src/main.ts`).
+
+On an uncached autoplay prime the raw payload is emitted here and reported to the controller with
+`notePlainSubtitleEmitted`, so the controller skips its own plain emit for that line and the
+overlay receives one plain payload followed by the annotated one.
+
+The pause is released by the controller's `onProcessingSettled` callback, which fires once it has
+no work left. Emits do not release it: the first emit for an uncached line is the plain payload
+that precedes tokenization, and a run can finish without emitting at all (a suppressed duplicate,
+a failed tokenization). Both controller methods return whether processing is now pending, and the
+caller resumes immediately when it is not — a repeated subtitle schedules no work, so no settle is
+coming and prefetching would otherwise idle for the rest of the cue.
+
+## Live Cue Delivery
+
+- A tokenization cache miss emits the plain cue synchronously. Tokenization remains serialized so
+  live work does not contend for Yomitan state.
+- If a newer cue arrives while an older line is still tokenizing, the newer plain cue or empty
+  clear payload is emitted immediately. The older tokenization result is dropped before it can
+  replace the current cue.
+- The current cue upgrades in place when its tokens and annotations are ready. This can reflow text
+  or character images, but cue visibility does not wait for that work.
 
 ## Emitted State
 
-- `emitSubtitle(payload)` maps to `emitSubtitlePayload(payload)`, which sends the normal
+- `emitSubtitle(payload)` maps to `emitSubtitlePayload(payload)`. Overlay windows and annotation
-  annotated subtitle payload to overlay windows and subtitle websocket listeners.
+  websocket listeners receive both the immediate plain cue and its later annotation upgrade.
+- The basic subtitle websocket receives the immediate plain cue only. Because its serialized
+  payload discards annotations, the later upgrade would be an identical duplicate and is skipped
+  when text and cue timing match.
 - Secondary priming reads mpv `secondary-sub-text`, stores it in
   `mpvClient.currentSecondarySubText`, and broadcasts `secondary-subtitle:set` to overlay windows.
 - If secondary `requestProperty` fails, the primary flow stays complete and only a debug line is

launcher/mpv.test.ts

@@ -222,7 +222,7 @@ test('buildMpvEnv preserves native Wayland env for supported Hyprland and Sway a
   });
 });
 
-test('buildMpvBackendArgs forces an explicit X11 renderer stack when backend resolves to x11', () => {
+test('buildMpvBackendArgs pins the X11 window context when backend resolves to x11', () => {
   withPlatform('linux', () => {
     assert.deepEqual(
       buildMpvBackendArgs(makeArgs({ backend: 'x11' }), {
@@ -230,12 +230,12 @@ test('buildMpvBackendArgs forces an explicit X11 renderer stack when backend res
         WAYLAND_DISPLAY: 'wayland-0',
         XDG_SESSION_TYPE: 'wayland',
       }),
-      ['--vo=gpu', '--gpu-api=opengl', '--gpu-context=x11egl,x11'],
+      ['--gpu-context=x11vk,x11egl,x11'],
     );
   });
 });
 
-test('buildMpvBackendArgs forces the same X11 renderer stack for unsupported Wayland auto fallback', () => {
+test('buildMpvBackendArgs pins the same X11 window context for unsupported Wayland auto fallback', () => {
   withPlatform('linux', () => {
     assert.deepEqual(
       buildMpvBackendArgs(makeArgs({ backend: 'auto' }), {
@@ -245,7 +245,7 @@ test('buildMpvBackendArgs forces the same X11 renderer stack for unsupported Way
         XDG_CURRENT_DESKTOP: 'KDE',
         XDG_SESSION_DESKTOP: 'plasma',
       }),
-      ['--vo=gpu', '--gpu-api=opengl', '--gpu-context=x11egl,x11'],
+      ['--gpu-context=x11vk,x11egl,x11'],
     );
   });
 });
@@ -292,9 +292,7 @@ test('buildConfiguredMpvDefaultArgs appends maximized launch mode to configured
         '--secondary-sub-visibility=no',
         '--alang=ja,jp,jpn,japanese,en,eng,english,enus,en-us',
         '--slang=ja,jp,jpn,japanese,en,eng,english,enus,en-us',
-        '--vo=gpu',
+        '--gpu-context=x11vk,x11egl,x11',
-        '--gpu-api=opengl',
-        '--gpu-context=x11egl,x11',
         '--window-maximized=yes',
       ],
     );

launcher/sync/engine-merge.test.ts

@@ -80,6 +80,11 @@ test('merges remote-only sessions with catalog, lifetime, and rollups', () => {
         { headword: '食べる', word: '食べた', reading: 'たべた', count: 1 },
       ],
     });
+    withWritableDb(remotePath, (db) => {
+      db.prepare(
+        `UPDATE imm_videos SET anime_assignment_locked = 1 WHERE video_key = 'showb-e1'`,
+      ).run();
+    });
 
     const summary = mergeSnapshotIntoDb(localPath, remotePath);
     assert.equal(summary.sessionsMerged, 1);
@@ -126,6 +131,14 @@ test('merges remote-only sessions with catalog, lifetime, and rollups', () => {
         `SELECT video_id FROM imm_videos WHERE video_key = 'showb-e1'`,
       )?.video_id,
     );
+    assert.equal(
+      queryOne<{ locked: number }>(
+        localPath,
+        'SELECT anime_assignment_locked AS locked FROM imm_videos WHERE video_id = ?',
+        [mergedVideoId],
+      )?.locked,
+      1,
+    );
     assert.equal(
       count(localPath, 'SELECT COUNT(*) AS n FROM imm_daily_rollups WHERE video_id = ?', [
         mergedVideoId,

launcher/test-support/immersion-db-schema.ts

@@ -1,4 +1,4 @@
-// Schema-version-18 shape of the tables the sync merge touches (plus the
+// Current schema shape of the tables the sync merge touches (plus the
 // app's indexes), mirroring ensureSchema / ensureLifetimeSummaryTables /
 // ensureStatsExcludedWordsTable in src/core/services/immersion-tracker/storage.ts.
 export const IMMERSION_DB_FIXTURE_DDL = `
@@ -39,6 +39,7 @@ export const IMMERSION_DB_FIXTURE_DDL = `
     parser_source TEXT,
     parser_confidence REAL,
     parse_metadata_json TEXT,
+    anime_assignment_locked INTEGER NOT NULL DEFAULT 0 CHECK(anime_assignment_locked IN (0, 1)),
     watched INTEGER NOT NULL DEFAULT 0,
     duration_ms INTEGER NOT NULL CHECK(duration_ms>=0),
     file_size_bytes INTEGER CHECK(file_size_bytes>=0),

package.json

@@ -2,7 +2,7 @@
   "name": "subminer",
   "productName": "SubMiner",
   "desktopName": "SubMiner.desktop",
-  "version": "0.19.0",
+  "version": "0.19.3",
   "description": "All-in-one sentence mining overlay with AnkiConnect and dictionary integration",
   "packageManager": "bun@1.3.5",
   "main": "dist/main-entry.js",
@@ -84,16 +84,18 @@
   "overrides": {
     "@xmldom/xmldom": "0.8.13",
     "app-builder-lib": "26.15.3",
-    "brace-expansion": "5.0.8",
+    "brace-expansion": "5.0.9",
     "electron-builder-squirrel-windows": "26.15.3",
+    "fast-uri": "3.1.5",
     "form-data": "4.0.6",
     "ip-address": "10.2.0",
-    "js-yaml": "4.3.0",
+    "js-yaml": "4.3.1",
     "lodash": "4.18.0",
     "minimatch": "10.2.5",
     "picomatch": "4.0.4",
     "tar": "7.5.21",
-    "tmp": "0.2.7"
+    "tmp": "0.2.7",
+    "undici": "7.29.0"
   },
   "keywords": [
     "anki",
@@ -125,7 +127,7 @@
     "@types/ws": "^8.18.1",
     "electron": "42.6.0",
     "electron-builder": "26.15.3",
-    "undici": "7.28.0",
+    "undici": "7.29.0",
     "esbuild": "^0.25.12",
     "eslint": "^10.8.0",
     "prettier": "^3.8.1",
@@ -258,6 +260,10 @@
       {
         "from": "dist/launcher/subminer",
         "to": "launcher/subminer"
+      },
+      {
+        "from": "CHANGELOG.md",
+        "to": "CHANGELOG.md"
       }
     ]
   },

release/release-notes.md

@@ -1,57 +1,38 @@
 ## Highlights
 ### Added
-- **Anki Maturity Known-Word Highlighting**
+- **In-App Changelog**
-  - Subtitle words you already know can now be color-coded by their Anki card maturity (new, learning, young, mature), like asbplayer's known-word coloring.
+  - View release notes without leaving the app, from the tray menu ("View Changelog") or the "What's New" button on update notifications.
-  - Enable it with `ankiConnect.knownWords.maturityEnabled` (or toggle it mid-session); tier colors and the "mature" day threshold are configurable, and the in-session help legend shows the active colors.
+  - Shows notes for the latest published release even when it's newer than your installed build, and falls back to the notes bundled with your install if the download fails.
-- **Cross-Machine Sync for Stats & Watch History**
+  - Older versions fold automatically, your installed version is badged, and newer ones are tagged "New"; navigate with `J`/`K` or the arrow keys, `Enter` to expand/collapse, `R` to refresh, and `Esc` to close.
-  - Sync immersion stats and watch history between machines over SSH from a new Sync window (tray menu → Sync Stats & History, or `subminer sync --ui`) or the CLI (`subminer sync <host>`).
-  - Save multiple devices with per-host sync direction, run one-click syncs with live progress, and take manual database snapshots for backup or transfer.
-  - Windows remotes are supported over OpenSSH, and hosts can auto-sync in the background on a schedule, even during playback.
-- **History Menu After Playback**
-  - After a watch-history episode ends (or mpv closes), the fzf/rofi launcher now offers to play the previous or next episode, rewatch, pick another episode, or quit, right from where you left off. Previous/Next continue across season folders.
-- **Delete Entire Library Titles from Stats**
-  - The stats Library detail view now has a "Delete Entry" action that removes a whole title in one step, episodes, sessions, subtitle lines, rollups, cover art, and vocabulary counts, instead of clearing it episode by episode.
-  - Delete progress (sessions, episodes, or whole titles) now shows app-wide with a progress bar and status toast visible from any tab or window.
-- **TsukiHime English Subtitle Downloads**
-  - Download subtitles for the currently playing video directly from TsukiHime, with Japanese loaded as the primary track and your configured secondary language alongside it.
 
 ### Changed
-- **Configurable Clipboard-Video Shortcut**
+- **Faster Subtitle Tokenization**
-  - The "append clipboard video to queue" shortcut is now configurable via `shortcuts.appendClipboardVideoToQueue` instead of fixed.
+  - Subtitle lines are parsed and looked up roughly twice as efficiently, with results cached across lines so repeated words and grammar no longer re-query the dictionary.
+  - Enabling a character dictionary no longer slows subtitle scanning as much, since name lookups now only check positions where a known name can actually start.
+  - Fixed related accuracy issues along the way: readings that could go missing on certain word endings, subtitle text that stayed unannotated after mining a card, character names that could drop out of disambiguation rules, and halfwidth-katakana character names that weren't recognized or read correctly.
 
 ### Fixed
-- **AniList Season Matching**
+- **Large Character Dictionary Generation**
-  - Season 2+ episodes now resolve to the correct AniList entry instead of silently falling back to season 1. SubMiner follows AniList's sequel relations to find the right season, and cover art and watch progress now use the same season-aware match.
+  - Big character dictionaries (long-running series like One Piece) no longer fail to install with a timeout error; the import time budget now scales with dictionary size instead of using a fixed 7-second limit.
-  - If a season still can't be found, SubMiner no longer force-writes progress or a cover to the season 1 entry, it skips the update and points you to a manual AniList override, which now fixes the character dictionary and watch progress together.
+  - The "Generating character dictionary" notification now shows real progress (character/page counts, image download progress with an ETA, name-processing progress) and an elapsed-time clock, so a long-running import no longer looks frozen.
-  - Manual overrides now stay applied consistently across every episode in a season folder, even when filenames guess differently episode to episode.
+- **Stats Deletion Responsiveness**
-- **Subtitle Highlighting Accuracy**
+  - Deleting sessions, episodes, or library entries on the stats page no longer freezes the page or an active video player; deletes are now batched into a single transaction.
-  - Fixed several known-word/annotation edge cases: part-of-speech exclusions now apply consistently to merged quote-particle tokens, annotations for rarer kanji are preserved, katakana punctuation is no longer mistaken for plain kana, and a specific noun-tagging case no longer loses its known+1 highlight.
+- **Subtitle Sidebar Clutter from Styled Subtitles**
-- **AnkiConnect Proxy Port Conflicts**
+  - Heavily typeset subtitles (karaoke openings/endings, stylized signs) no longer flood the subtitle sidebar with garbled vector-drawing text or duplicate "shadow" copies of the same line.
-  - Fixed a crash on video startup when another process already held the configured AnkiConnect proxy port; you'll now get a notification explaining how to resolve it instead.
+  - Subtitle text is now decoded consistently in one place, matching what mpv actually renders on screen, so it can no longer diverge or get cached inconsistently.
-- **AppImage Crash Notification on Quit**
+- **X11/XWayland Playback and Overlay Fixes**
-  - Fixed a spurious "Service Crash" desktop notification appearing after closing a video when running the Linux AppImage.
+  - Fixed a crash on the first fullscreen toggle when using an mpv `gpu-next` shader (e.g. ArtCNN) in X11/XWayland mode; SubMiner no longer forces mpv onto its older OpenGL renderer.
-- **Startup Playback Pause Timing**
+  - Fixed the overlay appearing oversized and offset from the video under fractional or mixed-monitor display scaling in X11/XWayland mode.
-  - Fixed playback occasionally resuming a couple seconds before subtitle tokenization actually finished warming up, most noticeable when resuming mid-episode or when a subtitle appears in the first two seconds.
-- **Stats Library Cover After Relinking**
-  - Fixed the stats Library grid showing a stale cover image after relinking a title to a different AniList entry.
-- **Faster Stats Deletes and Vocabulary Tab**
-  - Deleting sessions, episodes, and titles from stats is now dramatically faster and no longer stalls playback while it runs; the Vocabulary tab also loads much faster.
-  - The first launch after updating runs a one-time database migration (a few seconds, database grows about 20%); no action needed.
-- **Settings Validation and Stats Server Hardening**
-  - Invalid AnkiConnect settings now fall back safely with a warning instead of silently breaking, and the stats server is hardened against malformed requests, stalled AniList searches, and other edge cases that could previously crash it.
-- **Rofi Prompt Spacing**
-  - Fixed rofi menu prompts running into the search placeholder text with no space between them.
 
 ## What's Changed
 
-- feat(shortcuts): make clipboard-video-append shortcut configurable by @ksyasuda in #158
+- perf(tokenizer): single-pass Yomitan scan with cross-line caching and prefetch fixes by @ksyasuda in #185
-- refactor(tokenizer): extract subtitle annotation filter into rule table by @ksyasuda in #162
+- fix(subtitles): collapse duplicate ASS events and decode text once by @ksyasuda in #186
-- refactor(tsukihime): swap Animetosho backend for TsukiHime API by @ksyasuda in #165
+- feat(overlay): add in-app changelog modal by @ksyasuda in #187
-- refactor: split anki-connect and stats-server resolvers into modules by @ksyasuda in #169
+- fix(playback): stop forcing legacy OpenGL renderer on X11 mpv backend by @ksyasuda in #188
-- feat(launcher): add post-playback history menu with previous episode by @ksyasuda in #170
+- fix(dictionary): stop large character dictionaries from timing out by @ksyasuda in #189
-- Anki maturity-based known-word highlighting by @ksyasuda in #172
+- fix(overlay): handle X11 display scaling across monitors by @ksyasuda in #193
-- fix(anilist): resolve later seasons via sequel relations, not title guessing by @ksyasuda in #173
+- fix(stats): batch deletes off the main thread by @ksyasuda in #194
-- feat(stats): add library entry deletion and app-wide delete progress by @ksyasuda in #174
 
 ## Installation
 

src/anki-integration.ts

@@ -22,10 +22,12 @@ import { MediaGenerator } from './media-generator';
 import path from 'path';
 import {
   AnkiConnectConfig,
+  type CardKind,
   KikuDuplicateCardInfo,
   KikuFieldGroupingChoice,
   KikuMergePreviewResponse,
   NotificationOptions,
+  type WordCardKind,
 } from './types/anki';
 import { AiConfig } from './types/integrations';
 import type { KnownWordMaturityTier } from './types/subtitle';
@@ -50,6 +52,7 @@ import {
   withUpdateProgress,
   UiFeedbackState,
 } from './anki-integration/ui-feedback';
+import { applyCardKindFlagFields, resolveWordCardKindSetting } from './anki-integration/card-kinds';
 import { KnownWordCacheManager } from './anki-integration/known-word-cache';
 import { PollingRunner } from './anki-integration/polling';
 import type { AnkiConnectProxyServer } from './anki-integration/anki-connect-proxy';
@@ -83,8 +86,6 @@ interface NoteInfo {
   fields: Record<string, { value: string }>;
 }
 
-type CardKind = 'sentence' | 'audio' | 'word-and-sentence';
-
 function trimToNonEmptyString(value: unknown): string | null {
   if (typeof value !== 'string') return null;
   const trimmed = value.trim();
@@ -840,6 +841,7 @@ export class AnkiIntegration {
     kikuEnabled: boolean;
     kikuFieldGrouping: 'auto' | 'manual' | 'disabled';
     kikuDeleteDuplicateInAuto: boolean;
+    wordCardKind: WordCardKind;
   } {
     const lapis = this.getLapisConfig();
     const kiku = this.getKikuConfig();
@@ -852,6 +854,7 @@ export class AnkiIntegration {
       kikuEnabled: kiku.enabled,
       kikuFieldGrouping: (kiku.fieldGrouping || 'disabled') as 'auto' | 'manual' | 'disabled',
       kikuDeleteDuplicateInAuto: kiku.deleteDuplicateInAuto !== false,
+      wordCardKind: resolveWordCardKindSetting(this.config.lapisKiku?.wordCardKind),
     };
   }
 
@@ -1315,79 +1318,9 @@ export class AnkiIntegration {
     availableFieldNames: string[],
     cardKind: CardKind,
   ): void {
-    const audioFlagNames = ['IsAudioCard'];
+    applyCardKindFlagFields(updatedFields, cardKind, (preferredName) =>
-
+      this.resolveFieldName(availableFieldNames, preferredName),
-    if (cardKind === 'word-and-sentence') {
-      const wordAndSentenceFlag = this.resolveFieldName(
-        availableFieldNames,
-        'IsWordAndSentenceCard',
-      );
-      if (!wordAndSentenceFlag) {
-        return;
-      }
-      updatedFields[wordAndSentenceFlag] = 'x';
-
-      const sentenceFlag = this.resolveFieldName(availableFieldNames, 'IsSentenceCard');
-      if (sentenceFlag && sentenceFlag !== wordAndSentenceFlag) {
-        updatedFields[sentenceFlag] = '';
-      }
-
-      for (const audioFlagName of audioFlagNames) {
-        const resolved = this.resolveFieldName(availableFieldNames, audioFlagName);
-        if (resolved && resolved !== wordAndSentenceFlag) {
-          updatedFields[resolved] = '';
-        }
-      }
-      return;
-    }
-
-    if (cardKind === 'sentence') {
-      const sentenceFlag = this.resolveFieldName(availableFieldNames, 'IsSentenceCard');
-      if (sentenceFlag) {
-        updatedFields[sentenceFlag] = 'x';
-      }
-
-      for (const audioFlagName of audioFlagNames) {
-        const resolved = this.resolveFieldName(availableFieldNames, audioFlagName);
-        if (resolved && resolved !== sentenceFlag) {
-          updatedFields[resolved] = '';
-        }
-      }
-
-      const wordAndSentenceFlag = this.resolveFieldName(
-        availableFieldNames,
-        'IsWordAndSentenceCard',
-      );
-      if (wordAndSentenceFlag && wordAndSentenceFlag !== sentenceFlag) {
-        updatedFields[wordAndSentenceFlag] = '';
-      }
-      return;
-    }
-
-    const resolvedAudioFlags = Array.from(
-      new Set(
-        audioFlagNames
-          .map((name) => this.resolveFieldName(availableFieldNames, name))
-          .filter((name): name is string => Boolean(name)),
-      ),
     );
-    const audioFlagName = resolvedAudioFlags[0] || null;
-    if (audioFlagName) {
-      updatedFields[audioFlagName] = 'x';
-    }
-    for (const extraAudioFlag of resolvedAudioFlags.slice(1)) {
-      updatedFields[extraAudioFlag] = '';
-    }
-
-    const sentenceFlag = this.resolveFieldName(availableFieldNames, 'IsSentenceCard');
-    if (sentenceFlag && sentenceFlag !== audioFlagName) {
-      updatedFields[sentenceFlag] = '';
-    }
-
-    const wordAndSentenceFlag = this.resolveFieldName(availableFieldNames, 'IsWordAndSentenceCard');
-    if (wordAndSentenceFlag && wordAndSentenceFlag !== audioFlagName) {
-      updatedFields[wordAndSentenceFlag] = '';
-    }
   }
 
   private async showNotification(

src/anki-integration/card-creation-manual-update.test.ts

@@ -4,29 +4,23 @@ import test from 'node:test';
 import { CardCreationService } from './card-creation';
 import { toMpvEdlValue } from './mpv-edl-test-utils';
 import type { MediaInput } from '../media-generator';
-import type { AnkiConnectConfig } from '../types/anki';
+import type { AnkiConnectConfig, CardKind } from '../types/anki';
+import { applyCardKindFlagFields } from './card-kinds';
 
 type CardCreationDeps = ConstructorParameters<typeof CardCreationService>[0];
 
-function setWordAndSentenceCardTypeFields(
+function setCardTypeFields(
   updatedFields: Record<string, string>,
   availableFieldNames: string[],
-  cardKind: 'sentence' | 'audio' | 'word-and-sentence',
+  cardKind: CardKind,
 ): void {
-  if (cardKind !== 'word-and-sentence') return;
+  applyCardKindFlagFields(
-
+    updatedFields,
-  const resolveFieldName = (preferredName: string): string | null =>
+    cardKind,
-    availableFieldNames.find((name) => name.toLowerCase() === preferredName.toLowerCase()) ?? null;
+    (preferredName) =>
-  const wordAndSentenceFlag = resolveFieldName('IsWordAndSentenceCard');
+      availableFieldNames.find((name) => name.toLowerCase() === preferredName.toLowerCase()) ??
-  if (!wordAndSentenceFlag) return;
+      null,
-
+  );
-  updatedFields[wordAndSentenceFlag] = 'x';
-  for (const flagName of ['IsSentenceCard', 'IsAudioCard']) {
-    const resolved = resolveFieldName(flagName);
-    if (resolved && resolved !== wordAndSentenceFlag) {
-      updatedFields[resolved] = '';
-    }
-  }
 }
 
 function createManualUpdateService(overrides: Partial<CardCreationDeps> = {}): {
@@ -217,7 +211,7 @@ test('manual clipboard subtitle update marks Kiku word cards as word-and-sentenc
       kikuFieldGrouping: 'disabled',
       kikuDeleteDuplicateInAuto: false,
     }),
-    setCardTypeFields: setWordAndSentenceCardTypeFields,
+    setCardTypeFields,
   });
 
   await service.updateLastAddedFromClipboard('字幕');

src/anki-integration/card-creation.ts

@@ -3,7 +3,7 @@ import {
   getConfiguredWordFieldName,
   getPreferredWordValueFromExtractedFields,
 } from '../anki-field-config';
-import { AnkiConnectConfig } from '../types/anki';
+import { AnkiConnectConfig, type CardKind, type WordCardKind } from '../types/anki';
 import { createLogger } from '../logger';
 import type { MediaInput } from '../media-input';
 import { SubtitleTimingTracker } from '../subtitle-timing-tracker';
@@ -15,7 +15,7 @@ import {
   resolveAudioStreamIndexForMediaGeneration,
   type MediaGenerationInputResolverOptions,
 } from './media-source';
-import { shouldMarkWordAndSentenceCard } from './note-field-utils';
+import { resolveWordCardKind } from './note-field-utils';
 import type { PendingYoutubeMediaUpdate } from './pending-youtube-media';
 import { resolveMpvVolumeScale } from './mpv-volume';
 
@@ -42,8 +42,6 @@ export interface CardCreationNoteInfo {
   fields: Record<string, { value: string }>;
 }
 
-type CardKind = 'sentence' | 'audio' | 'word-and-sentence';
-
 interface CardCreationClient {
   addNote(
     deck: string,
@@ -136,6 +134,7 @@ interface CardCreationDeps {
     kikuEnabled: boolean;
     kikuFieldGrouping: 'auto' | 'manual' | 'disabled';
     kikuDeleteDuplicateInAuto: boolean;
+    wordCardKind?: WordCardKind;
   };
   getFallbackDurationSeconds: () => number;
   appendKnownWordsFromNoteInfo: (noteInfo: CardCreationNoteInfo) => void;
@@ -274,12 +273,9 @@ export class CardCreationService {
         if (sentenceField) {
           const processedSentence = this.deps.processSentence(sentence, fields);
           updatedFields[sentenceField] = processedSentence;
-          if (shouldMarkWordAndSentenceCard(noteInfo, sentenceCardConfig)) {
+          const wordCardKind = resolveWordCardKind(noteInfo, sentenceCardConfig);
-            this.deps.setCardTypeFields(
+          if (wordCardKind) {
-              updatedFields,
+            this.deps.setCardTypeFields(updatedFields, Object.keys(noteInfo.fields), wordCardKind);
-              Object.keys(noteInfo.fields),
-              'word-and-sentence',
-            );
           }
           updatePerformed = true;
         }

src/anki-integration/card-kinds.test.ts

@@ -0,0 +1,64 @@
+import assert from 'node:assert/strict';
+import test from 'node:test';
+
+import { applyCardKindFlagFields } from './card-kinds';
+
+function resolverFor(availableFieldNames: string[]) {
+  return (preferredName: string): string | null =>
+    availableFieldNames.find((name) => name.toLowerCase() === preferredName.toLowerCase()) ?? null;
+}
+
+const KIKU_FLAG_FIELDS = ['IsWordAndSentenceCard', 'IsClickCard', 'IsSentenceCard', 'IsAudioCard'];
+
+test('flags the requested card kind and clears the others', () => {
+  const fields: Record<string, string> = {};
+
+  applyCardKindFlagFields(fields, 'click', resolverFor(KIKU_FLAG_FIELDS));
+
+  assert.deepEqual(fields, {
+    IsClickCard: 'x',
+    IsWordAndSentenceCard: '',
+    IsSentenceCard: '',
+    IsAudioCard: '',
+  });
+});
+
+test('matches flag fields case-insensitively', () => {
+  const fields: Record<string, string> = {};
+
+  applyCardKindFlagFields(fields, 'word-and-sentence', resolverFor(['iswordandsentencecard']));
+
+  assert.deepEqual(fields, { iswordandsentencecard: 'x' });
+});
+
+test('leaves flags untouched when the note type has no flag for a word card kind', () => {
+  const fields: Record<string, string> = {};
+
+  applyCardKindFlagFields(
+    fields,
+    'click',
+    resolverFor(['IsWordAndSentenceCard', 'IsSentenceCard']),
+  );
+
+  assert.deepEqual(fields, {});
+});
+
+test('clears stale flags for explicit mine actions even without the target flag', () => {
+  const fields: Record<string, string> = {};
+
+  applyCardKindFlagFields(
+    fields,
+    'audio',
+    resolverFor(['IsWordAndSentenceCard', 'IsSentenceCard']),
+  );
+
+  assert.deepEqual(fields, { IsWordAndSentenceCard: '', IsSentenceCard: '' });
+});
+
+test('does not blank the target flag it just set', () => {
+  const fields: Record<string, string> = {};
+
+  applyCardKindFlagFields(fields, 'sentence', resolverFor(['IsSentenceCard']));
+
+  assert.deepEqual(fields, { IsSentenceCard: 'x' });
+});

src/anki-integration/card-kinds.ts

@@ -0,0 +1,63 @@
+import type { CardKind, WordCardKind } from '../types/anki';
+
+/**
+ * Kiku/Lapis note types decide which card a note generates from mutually exclusive
+ * `Is...Card` flag fields. Setting one always means clearing the others.
+ */
+export const CARD_KIND_FLAG_FIELDS: Record<CardKind, string> = {
+  'word-and-sentence': 'IsWordAndSentenceCard',
+  click: 'IsClickCard',
+  sentence: 'IsSentenceCard',
+  audio: 'IsAudioCard',
+};
+
+export const WORD_CARD_KINDS: readonly WordCardKind[] = [
+  'word-and-sentence',
+  'click',
+  'sentence',
+  'audio',
+  'none',
+];
+
+export const DEFAULT_WORD_CARD_KIND: WordCardKind = 'word-and-sentence';
+
+/**
+ * Card kinds SubMiner marks on its own initiative (word cards). They are only applied
+ * when the note type actually carries the matching flag field, so plain note types keep
+ * their fields untouched.
+ */
+const IMPLICIT_CARD_KINDS = new Set<CardKind>(['word-and-sentence', 'click']);
+
+export function isWordCardKind(value: unknown): value is WordCardKind {
+  return typeof value === 'string' && WORD_CARD_KINDS.includes(value as WordCardKind);
+}
+
+export function resolveWordCardKindSetting(value: unknown): WordCardKind {
+  return isWordCardKind(value) ? value : DEFAULT_WORD_CARD_KIND;
+}
+
+/**
+ * Flags `cardKind` on the note and clears every other card-kind flag it has, so the note
+ * never ends up claiming to be two kinds of card at once.
+ */
+export function applyCardKindFlagFields(
+  updatedFields: Record<string, string>,
+  cardKind: CardKind,
+  resolveFieldName: (preferredName: string) => string | null,
+): void {
+  const targetFlag = resolveFieldName(CARD_KIND_FLAG_FIELDS[cardKind]);
+  if (!targetFlag && IMPLICIT_CARD_KINDS.has(cardKind)) {
+    return;
+  }
+  if (targetFlag) {
+    updatedFields[targetFlag] = 'x';
+  }
+
+  for (const [kind, flagName] of Object.entries(CARD_KIND_FLAG_FIELDS)) {
+    if (kind === cardKind) continue;
+    const resolved = resolveFieldName(flagName);
+    if (resolved && resolved !== targetFlag) {
+      updatedFields[resolved] = '';
+    }
+  }
+}

src/anki-integration/note-field-utils.test.ts

@@ -0,0 +1,118 @@
+import assert from 'node:assert/strict';
+import test from 'node:test';
+
+import { resolveWordCardKind, type NoteFieldValueInfo } from './note-field-utils';
+
+function kikuNote(values: Record<string, string> = {}): NoteFieldValueInfo {
+  const defaults: Record<string, string> = {
+    Expression: '単語',
+    Sentence: '',
+    IsWordAndSentenceCard: '',
+    IsClickCard: '',
+    IsSentenceCard: '',
+    IsAudioCard: '',
+  };
+  return {
+    fields: Object.fromEntries(
+      Object.entries({ ...defaults, ...values }).map(([name, value]) => [name, { value }]),
+    ),
+  };
+}
+
+test('marks word-and-sentence cards by default when Kiku is enabled', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote(), { lapisEnabled: false, kikuEnabled: true }),
+    'word-and-sentence',
+  );
+});
+
+test('honors the configured word card kind', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote(), {
+      lapisEnabled: false,
+      kikuEnabled: true,
+      wordCardKind: 'click',
+    }),
+    'click',
+  );
+});
+
+test('marks nothing when neither Kiku nor Lapis is enabled', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote(), {
+      lapisEnabled: false,
+      kikuEnabled: false,
+      wordCardKind: 'click',
+    }),
+    null,
+  );
+});
+
+test('marks nothing when the word card kind is "none"', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote(), {
+      lapisEnabled: true,
+      kikuEnabled: false,
+      wordCardKind: 'none',
+    }),
+    null,
+  );
+});
+
+test('falls back to the default kind for an unrecognized setting', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote(), {
+      lapisEnabled: false,
+      kikuEnabled: true,
+      wordCardKind: 'bogus' as never,
+    }),
+    'word-and-sentence',
+  );
+});
+
+test('marks nothing when the note type lacks the configured flag field', () => {
+  const note: NoteFieldValueInfo = {
+    fields: { Expression: { value: '単語' }, Sentence: { value: '' } },
+  };
+
+  assert.equal(
+    resolveWordCardKind(note, { lapisEnabled: false, kikuEnabled: true, wordCardKind: 'click' }),
+    null,
+  );
+});
+
+test('leaves cards already mined as sentence or audio cards alone', () => {
+  for (const flagField of ['IsSentenceCard', 'IsAudioCard']) {
+    assert.equal(
+      resolveWordCardKind(kikuNote({ [flagField]: 'x' }), {
+        lapisEnabled: false,
+        kikuEnabled: true,
+        wordCardKind: 'click',
+      }),
+      null,
+      flagField,
+    );
+  }
+});
+
+test('re-affirms the configured kind when the note already carries its flag', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote({ IsSentenceCard: 'x' }), {
+      lapisEnabled: false,
+      kikuEnabled: true,
+      wordCardKind: 'sentence',
+    }),
+    'sentence',
+  );
+});
+
+test('overrides a differently flagged word card', () => {
+  assert.equal(
+    resolveWordCardKind(kikuNote({ IsWordAndSentenceCard: 'x' }), {
+      lapisEnabled: false,
+      kikuEnabled: true,
+      wordCardKind: 'click',
+    }),
+    'click',
+  );
+});

src/anki-integration/note-field-utils.ts

@@ -1,3 +1,13 @@
+import type { CardKind, WordCardKind } from '../types/anki';
+import { createLogger } from '../logger';
+import {
+  CARD_KIND_FLAG_FIELDS,
+  DEFAULT_WORD_CARD_KIND,
+  resolveWordCardKindSetting,
+} from './card-kinds';
+
+const log = createLogger('anki').child('integration.note-fields');
+
 export interface NoteFieldValueInfo {
   fields: Record<string, { value: string }>;
 }
@@ -16,22 +26,57 @@ export function hasNoteFieldValue(noteInfo: NoteFieldValueInfo, preferredName: s
   return (getNoteFieldValue(noteInfo, preferredName) ?? '').trim().length > 0;
 }
 
-export function shouldMarkWordAndSentenceCard(
+/** Flags set only by an explicit mine action; a note carrying one is not a word card. */
-  noteInfo: NoteFieldValueInfo,
+const EXPLICIT_CARD_FLAG_FIELDS = [CARD_KIND_FLAG_FIELDS.sentence, CARD_KIND_FLAG_FIELDS.audio];
-  sentenceCardConfig: { lapisEnabled: boolean; kikuEnabled: boolean },
-): boolean {
-  if (!sentenceCardConfig.lapisEnabled && !sentenceCardConfig.kikuEnabled) {
-    return false;
-  }
 
-  const wordAndSentenceValue = getNoteFieldValue(noteInfo, 'IsWordAndSentenceCard');
+const warnedMissingFlagFields = new Set<CardKind>();
-  if (wordAndSentenceValue === null) {
+
-    return false;
+function warnMissingFlagFieldOnce(wordCardKind: CardKind, flagField: string): void {
+  if (wordCardKind === DEFAULT_WORD_CARD_KIND || warnedMissingFlagFields.has(wordCardKind)) {
+    // The default kind is also the fallback for plain note types, so its absence is expected.
+    return;
   }
-  if (wordAndSentenceValue.trim().length > 0) {
+  warnedMissingFlagFields.add(wordCardKind);
-    return true;
+  log.warn(
-  }
+    `Word card type "${wordCardKind}" is configured but the note has no ${flagField} field; leaving card type flags unchanged.`,
-  return (
-    !hasNoteFieldValue(noteInfo, 'IsSentenceCard') && !hasNoteFieldValue(noteInfo, 'IsAudioCard')
   );
 }
+
+/**
+ * Card kind to flag when SubMiner fills a word card's sentence, or null to leave the
+ * card-kind flags alone. Kiku/Lapis only: other note types have no such fields.
+ */
+export function resolveWordCardKind(
+  noteInfo: NoteFieldValueInfo,
+  sentenceCardConfig: {
+    lapisEnabled: boolean;
+    kikuEnabled: boolean;
+    wordCardKind?: WordCardKind;
+  },
+): CardKind | null {
+  if (!sentenceCardConfig.lapisEnabled && !sentenceCardConfig.kikuEnabled) {
+    return null;
+  }
+
+  const wordCardKind = resolveWordCardKindSetting(sentenceCardConfig.wordCardKind);
+  if (wordCardKind === 'none') {
+    return null;
+  }
+
+  const flagField = CARD_KIND_FLAG_FIELDS[wordCardKind];
+  const flagValue = getNoteFieldValue(noteInfo, flagField);
+  if (flagValue === null) {
+    // Note type has no flag field for the configured kind.
+    warnMissingFlagFieldOnce(wordCardKind, flagField);
+    return null;
+  }
+  if (flagValue.trim().length > 0) {
+    return wordCardKind;
+  }
+
+  const alreadyExplicitCard = EXPLICIT_CARD_FLAG_FIELDS.some(
+    (fieldName) =>
+      fieldName.toLowerCase() !== flagField.toLowerCase() && hasNoteFieldValue(noteInfo, fieldName),
+  );
+  return alreadyExplicitCard ? null : wordCardKind;
+}

src/anki-integration/note-update-workflow.test.ts

@@ -6,26 +6,21 @@ import {
   type NoteUpdateWorkflowNoteInfo,
 } from './note-update-workflow';
 import type { SubtitleMiningContext } from '../types/subtitle';
+import type { CardKind } from '../types/anki';
+import { applyCardKindFlagFields } from './card-kinds';
 
-function setWordAndSentenceCardTypeFields(
+function setCardTypeFields(
   updatedFields: Record<string, string>,
   availableFieldNames: string[],
-  cardKind: 'word-and-sentence',
+  cardKind: CardKind,
 ): void {
-  assert.equal(cardKind, 'word-and-sentence');
+  applyCardKindFlagFields(
-  const resolveFieldName = (preferredName: string): string | null =>
+    updatedFields,
-    availableFieldNames.find((name) => name.toLowerCase() === preferredName.toLowerCase()) ?? null;
+    cardKind,
-
+    (preferredName) =>
-  const wordAndSentenceFlag = resolveFieldName('IsWordAndSentenceCard');
+      availableFieldNames.find((name) => name.toLowerCase() === preferredName.toLowerCase()) ??
-  if (!wordAndSentenceFlag) return;
+      null,
-
+  );
-  updatedFields[wordAndSentenceFlag] = 'x';
-  for (const flagName of ['IsSentenceCard', 'IsAudioCard']) {
-    const resolved = resolveFieldName(flagName);
-    if (resolved && resolved !== wordAndSentenceFlag) {
-      updatedFields[resolved] = '';
-    }
-  }
 }
 
 function createWorkflowHarness() {
@@ -79,7 +74,7 @@ function createWorkflowHarness() {
     handleFieldGroupingManual: async (_originalNoteId, _newNoteId, _newNoteInfo, _expression) =>
       false,
     processSentence: (text: string, _noteFields: Record<string, string>) => text,
-    setCardTypeFields: setWordAndSentenceCardTypeFields,
+    setCardTypeFields,
     resolveConfiguredFieldName: (noteInfo: NoteUpdateWorkflowNoteInfo, preferred?: string) => {
       if (!preferred) return null;
       const names = Object.keys(noteInfo.fields);
@@ -183,6 +178,73 @@ test('NoteUpdateWorkflow marks enriched Kiku word cards as word-and-sentence car
   });
 });
 
+test('NoteUpdateWorkflow marks the configured word card kind instead of word-and-sentence', async () => {
+  const harness = createWorkflowHarness();
+  harness.deps.getEffectiveSentenceCardConfig = () => ({
+    sentenceField: 'Sentence',
+    lapisEnabled: false,
+    kikuEnabled: true,
+    kikuFieldGrouping: 'manual',
+    wordCardKind: 'click',
+  });
+  harness.deps.client.notesInfo = async () =>
+    [
+      {
+        noteId: 42,
+        fields: {
+          Expression: { value: 'taberu' },
+          Sentence: { value: '' },
+          IsWordAndSentenceCard: { value: 'x' },
+          IsClickCard: { value: '' },
+          IsSentenceCard: { value: '' },
+          IsAudioCard: { value: '' },
+        },
+      },
+    ] satisfies NoteUpdateWorkflowNoteInfo[];
+
+  await harness.workflow.execute(42);
+
+  assert.equal(harness.updates.length, 1);
+  assert.deepEqual(harness.updates[0]?.fields, {
+    Sentence: 'subtitle-text',
+    IsClickCard: 'x',
+    IsWordAndSentenceCard: '',
+    IsSentenceCard: '',
+    IsAudioCard: '',
+  });
+});
+
+test('NoteUpdateWorkflow leaves card type flags alone when the word card kind is none', async () => {
+  const harness = createWorkflowHarness();
+  harness.deps.getEffectiveSentenceCardConfig = () => ({
+    sentenceField: 'Sentence',
+    lapisEnabled: false,
+    kikuEnabled: true,
+    kikuFieldGrouping: 'manual',
+    wordCardKind: 'none',
+  });
+  harness.deps.client.notesInfo = async () =>
+    [
+      {
+        noteId: 42,
+        fields: {
+          Expression: { value: 'taberu' },
+          Sentence: { value: '' },
+          IsWordAndSentenceCard: { value: '' },
+          IsSentenceCard: { value: '' },
+          IsAudioCard: { value: '' },
+        },
+      },
+    ] satisfies NoteUpdateWorkflowNoteInfo[];
+
+  await harness.workflow.execute(42);
+
+  assert.equal(harness.updates.length, 1);
+  assert.deepEqual(harness.updates[0]?.fields, {
+    Sentence: 'subtitle-text',
+  });
+});
+
 test('NoteUpdateWorkflow does not set Kiku card flags when Lapis and Kiku are disabled', async () => {
   const harness = createWorkflowHarness();
   harness.deps.client.notesInfo = async () =>

src/anki-integration/note-update-workflow.ts

@@ -1,7 +1,8 @@
 import { DEFAULT_ANKI_CONNECT_CONFIG } from '../config';
 import { getPreferredWordValueFromExtractedFields } from '../anki-field-config';
 import type { SubtitleMiningContext } from '../types/subtitle';
-import { shouldMarkWordAndSentenceCard } from './note-field-utils';
+import type { CardKind, WordCardKind } from '../types/anki';
+import { resolveWordCardKind } from './note-field-utils';
 
 export interface NoteUpdateWorkflowNoteInfo {
   noteId: number;
@@ -39,6 +40,7 @@ export interface NoteUpdateWorkflowDeps {
     lapisEnabled: boolean;
     kikuEnabled: boolean;
     kikuFieldGrouping: 'auto' | 'manual' | 'disabled';
+    wordCardKind?: WordCardKind;
   };
   appendKnownWordsFromNoteInfo: (noteInfo: NoteUpdateWorkflowNoteInfo) => void;
   extractFields: (fields: Record<string, { value: string }>) => Record<string, string>;
@@ -67,7 +69,7 @@ export interface NoteUpdateWorkflowDeps {
   setCardTypeFields: (
     updatedFields: Record<string, string>,
     availableFieldNames: string[],
-    cardKind: 'word-and-sentence',
+    cardKind: CardKind,
   ) => void;
   resolveConfiguredFieldName: (
     noteInfo: NoteUpdateWorkflowNoteInfo,
@@ -207,12 +209,9 @@ export class NoteUpdateWorkflow {
       if (sentenceField && currentSubtitleText) {
         const processedSentence = this.deps.processSentence(currentSubtitleText, fields);
         updatedFields[sentenceField] = processedSentence;
-        if (shouldMarkWordAndSentenceCard(noteInfo, sentenceCardConfig)) {
+        const wordCardKind = resolveWordCardKind(noteInfo, sentenceCardConfig);
-          this.deps.setCardTypeFields(
+        if (wordCardKind) {
-            updatedFields,
+          this.deps.setCardTypeFields(updatedFields, Object.keys(noteInfo.fields), wordCardKind);
-            Object.keys(noteInfo.fields),
-            'word-and-sentence',
-          );
         }
         updatePerformed = true;
       }

src/anki-integration/runtime.ts

@@ -116,6 +116,10 @@ export function normalizeAnkiIntegrationConfig(config: AnkiConnectConfig): AnkiC
       ...DEFAULT_ANKI_CONNECT_CONFIG.isKiku,
       ...(config.isKiku ?? {}),
     },
+    lapisKiku: {
+      ...DEFAULT_ANKI_CONNECT_CONFIG.lapisKiku,
+      ...(config.lapisKiku ?? {}),
+    },
   } as AnkiConnectConfig;
 }
 
@@ -205,6 +209,10 @@ export class AnkiIntegrationRuntime {
         patch.isKiku !== undefined
           ? { ...this.config.isKiku, ...patch.isKiku }
           : this.config.isKiku,
+      lapisKiku:
+        patch.lapisKiku !== undefined
+          ? { ...this.config.lapisKiku, ...patch.lapisKiku }
+          : this.config.lapisKiku,
     };
     this.config = normalizeAnkiIntegrationConfig(mergedConfig);
     this.deps.onConfigChanged?.(this.config);

src/config/config.test.ts

@@ -2738,6 +2738,43 @@ test('ignores deprecated isLapis sentence-card field overrides', () => {
   );
 });
 
+test('accepts a Kiku/Lapis word card kind and warns on an unknown one', () => {
+  const dir = makeTempDir();
+  fs.writeFileSync(
+    path.join(dir, 'config.jsonc'),
+    `{
+      "ankiConnect": {
+        "isKiku": { "enabled": true },
+        "lapisKiku": { "wordCardKind": "click" }
+      }
+    }`,
+    'utf-8',
+  );
+
+  const service = new ConfigService(dir);
+  assert.equal(service.getConfig().ankiConnect.lapisKiku.wordCardKind, 'click');
+  assert.equal(service.getWarnings().length, 0);
+
+  const invalidDir = makeTempDir();
+  fs.writeFileSync(
+    path.join(invalidDir, 'config.jsonc'),
+    `{
+      "ankiConnect": {
+        "lapisKiku": { "wordCardKind": "isClickCard" }
+      }
+    }`,
+    'utf-8',
+  );
+
+  const invalidService = new ConfigService(invalidDir);
+  assert.equal(invalidService.getConfig().ankiConnect.lapisKiku.wordCardKind, 'word-and-sentence');
+  assert.ok(
+    invalidService
+      .getWarnings()
+      .some((warning) => warning.path === 'ankiConnect.lapisKiku.wordCardKind'),
+  );
+});
+
 test('accepts valid ankiConnect knownWords deck object', () => {
   const dir = makeTempDir();
   fs.writeFileSync(

src/config/definitions/defaults-integrations.ts

@@ -91,6 +91,9 @@ export const INTEGRATIONS_DEFAULT_CONFIG: Pick<
       fieldGrouping: 'disabled',
       deleteDuplicateInAuto: true,
     },
+    lapisKiku: {
+      wordCardKind: 'word-and-sentence',
+    },
   },
   jimaku: {
     apiBaseUrl: 'https://jimaku.cc',

src/config/definitions/options-integrations.ts

@@ -1,4 +1,5 @@
 import { ResolvedConfig } from '../../types/config';
+import { WORD_CARD_KINDS } from '../../anki-integration/card-kinds';
 import { MPV_LAUNCH_MODE_VALUES } from '../../shared/mpv-launch-mode';
 import {
   NOTIFICATION_TYPE_VALUES,
@@ -374,6 +375,21 @@ export function buildIntegrationConfigOptionRegistry(
       defaultValue: defaultConfig.ankiConnect.isLapis.sentenceCardModel,
       description: 'Note type name used by Lapis sentence cards.',
     },
+    {
+      path: 'ankiConnect.lapisKiku.wordCardKind',
+      kind: 'enum',
+      enumValues: WORD_CARD_KINDS,
+      enumLabels: {
+        'word-and-sentence': 'Word and sentence card (IsWordAndSentenceCard)',
+        click: 'Click card (IsClickCard)',
+        sentence: 'Sentence card (IsSentenceCard)',
+        audio: 'Audio card (IsAudioCard)',
+        none: 'Leave card type flags untouched',
+      },
+      defaultValue: defaultConfig.ankiConnect.lapisKiku.wordCardKind,
+      description:
+        'Card-type flag SubMiner marks on Kiku/Lapis word cards. Only one flag is set at a time; the others are cleared. Requires isKiku.enabled or isLapis.enabled.',
+    },
     {
       path: 'ankiConnect.metadata.pattern',
       kind: 'string',

src/config/definitions/template-sections.ts

@@ -135,7 +135,7 @@ const INTEGRATION_TEMPLATE_SECTIONS: ConfigTemplateSection[] = [
     title: 'AnkiConnect Integration',
     description: ['Automatic Anki updates and media generation options.'],
     notes: [
-      'Hot-reload: ankiConnect.ai.enabled, media.normalizeAudio/mirrorMpvVolume, knownWords, nPlusOne, fields.word/audio/image/sentence/miscInfo, behavior.autoUpdateNewCards, isLapis.sentenceCardModel, and isKiku.fieldGrouping update live while SubMiner is running.',
+      'Hot-reload: ankiConnect.ai.enabled, media.normalizeAudio/mirrorMpvVolume, knownWords, nPlusOne, fields.word/audio/image/sentence/miscInfo, behavior.autoUpdateNewCards, isLapis.sentenceCardModel, isKiku.fieldGrouping, and lapisKiku.wordCardKind update live while SubMiner is running.',
       'Shared AI provider transport settings are read from top-level ai and typically require restart.',
       'Most other AnkiConnect settings still require restart.',
     ],

src/config/resolve/anki-connect.ts

@@ -1,6 +1,7 @@
 import type { ResolveContext } from './context';
 import { initializeAnkiConnectResolution } from './anki-connect/initialize';
 import { applyAnkiKikuResolution } from './anki-connect/kiku';
+import { applyAnkiLapisKikuResolution } from './anki-connect/lapis-kiku';
 import { applyAnkiKnownWordsResolution } from './anki-connect/known-words';
 import { applyAnkiLegacyResolution } from './anki-connect/legacy';
 import { applyAnkiModernResolution } from './anki-connect/modern';
@@ -22,4 +23,5 @@ export function applyAnkiConnectResolution(context: ResolveContext): void {
   applyAnkiLegacyResolution(context, ankiConnect, behavior, fields, media, metadata);
   applyAnkiKnownWordsResolution(context, ankiConnect, behavior);
   applyAnkiKikuResolution(context);
+  applyAnkiLapisKikuResolution(context, ankiConnect);
 }

src/config/resolve/anki-connect/initialize.ts

@@ -77,5 +77,8 @@ export function initializeAnkiConnectResolution(
         ? (ankiConnect.isKiku as (typeof context.resolved)['ankiConnect']['isKiku'])
         : {}),
     },
+    lapisKiku: {
+      ...context.resolved.ankiConnect.lapisKiku,
+    },
   };
 }

src/config/resolve/anki-connect/lapis-kiku.ts

@@ -0,0 +1,39 @@
+import { isWordCardKind, WORD_CARD_KINDS } from '../../../anki-integration/card-kinds';
+import { DEFAULT_CONFIG } from '../../definitions';
+import type { ResolveContext } from '../context';
+import { isObject } from '../shared';
+
+export function applyAnkiLapisKikuResolution(
+  context: ResolveContext,
+  ankiConnect: Record<string, unknown>,
+): void {
+  if (!isObject(ankiConnect.lapisKiku)) {
+    if (ankiConnect.lapisKiku !== undefined) {
+      context.warn(
+        'ankiConnect.lapisKiku',
+        ankiConnect.lapisKiku,
+        context.resolved.ankiConnect.lapisKiku,
+        'Expected object.',
+      );
+    }
+    return;
+  }
+
+  const wordCardKind = ankiConnect.lapisKiku.wordCardKind;
+  if (wordCardKind === undefined) {
+    return;
+  }
+  if (isWordCardKind(wordCardKind)) {
+    context.resolved.ankiConnect.lapisKiku.wordCardKind = wordCardKind;
+    return;
+  }
+
+  context.warn(
+    'ankiConnect.lapisKiku.wordCardKind',
+    wordCardKind,
+    DEFAULT_CONFIG.ankiConnect.lapisKiku.wordCardKind,
+    `Expected one of ${WORD_CARD_KINDS.join(', ')}.`,
+  );
+  context.resolved.ankiConnect.lapisKiku.wordCardKind =
+    DEFAULT_CONFIG.ankiConnect.lapisKiku.wordCardKind;
+}

src/config/settings/registry.ts

@@ -221,6 +221,7 @@ const LABEL_OVERRIDES: Record<string, string> = {
   'ankiConnect.nPlusOne.enabled': 'Enabled',
   'ankiConnect.isLapis.enabled': 'Enable Lapis Features',
   'ankiConnect.isKiku.enabled': 'Enable Kiku Features',
+  'ankiConnect.lapisKiku.wordCardKind': 'Word Card Type',
   'stats.toggleKey': 'Toggle Stats Overlay',
   'shortcuts.openCharacterDictionaryManager': 'Open Character Dictionary Manager',
   'subtitleSidebar.pauseVideoOnHover': 'Pause Video On Hover - Sidebar',
@@ -255,6 +256,8 @@ const DESCRIPTION_OVERRIDES: Record<string, string> = {
     'Enable Lapis-specific mining behavior and sentence-card model targeting. When Kiku is enabled, Lapis features still work and Kiku-specific features are added on top.',
   'ankiConnect.isLapis.sentenceCardModel':
     'Anki note type used for Lapis sentence cards. Select from note types reported by AnkiConnect.',
+  'ankiConnect.lapisKiku.wordCardKind':
+    'Card-type flag marked on mined word cards. Only one flag is set at a time; the others are cleared. Requires Kiku or Lapis to be enabled.',
   'subtitleStyle.css':
     'CSS declarations applied to primary subtitles. Includes color, background-color, and all font properties.',
   'subtitleStyle.secondary.css':
@@ -401,7 +404,11 @@ function categoryAndSection(path: string): { category: ConfigSettingsCategory; s
   if (path.startsWith('ankiConnect.media.')) {
     return { category: 'mining-anki', section: 'Media Capture' };
   }
-  if (path.startsWith('ankiConnect.isKiku.') || path.startsWith('ankiConnect.isLapis.')) {
+  if (
+    path.startsWith('ankiConnect.isKiku.') ||
+    path.startsWith('ankiConnect.isLapis.') ||
+    path.startsWith('ankiConnect.lapisKiku.')
+  ) {
     return { category: 'mining-anki', section: 'Kiku/Lapis Features' };
   }
   if (path.startsWith('ankiConnect.ai.')) {
@@ -702,6 +709,7 @@ function restartBehaviorForPath(path: string): ConfigSettingsRestartBehavior {
     path === 'ankiConnect.fields.miscInfo' ||
     path === 'ankiConnect.isLapis.sentenceCardModel' ||
     path === 'ankiConnect.isKiku.fieldGrouping' ||
+    path === 'ankiConnect.lapisKiku.wordCardKind' ||
     path === 'mpv.aniskipEnabled' ||
     path === 'mpv.aniskipButtonKey' ||
     path === 'stats.toggleKey' ||

src/core/services/__tests__/immersion-tracker-write-queue.test.ts

@@ -0,0 +1,322 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import type { DatabaseSync } from '../immersion-tracker/sqlite';
+
+type ImmersionTrackerService = import('../immersion-tracker-service').ImmersionTrackerService;
+type ImmersionTrackerServiceCtor =
+  typeof import('../immersion-tracker-service').ImmersionTrackerService;
+
+let trackerCtor: ImmersionTrackerServiceCtor | null = null;
+
+async function loadTrackerCtor(): Promise<ImmersionTrackerServiceCtor> {
+  if (trackerCtor) return trackerCtor;
+  const mod = await import('../immersion-tracker-service');
+  trackerCtor = mod.ImmersionTrackerService;
+  return trackerCtor;
+}
+
+function makeDbPath(): string {
+  const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'subminer-write-queue-test-'));
+  return path.join(dir, 'immersion.sqlite');
+}
+
+function cleanupDbPath(dbPath: string): void {
+  const dir = path.dirname(dbPath);
+  if (!fs.existsSync(dir)) return;
+  fs.rmSync(dir, { recursive: true, force: true });
+}
+
+interface TrackerInternals {
+  db: DatabaseSync;
+  queue: unknown[];
+  recordWrite: (write: Record<string, unknown>) => void;
+  deleteSession: (sessionId: number) => Promise<void>;
+  mergeAnime: (targetAnimeId: number, sourceAnimeIds: number[]) => Promise<unknown>;
+  moveVideoToAnime: (videoId: number, targetAnimeId: number) => Promise<unknown>;
+  rebuildLifetimeSummaries: () => Promise<unknown>;
+  reassignAnimeAnilist: (animeId: number, info: { anilistId: number }) => Promise<void>;
+  flushNow: () => void;
+  writeLock: { locked: boolean };
+}
+
+test('delete maintenance fails closed when queued writes cannot drain', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  let deleteRunnerCalls = 0;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor(
+      { dbPath, policy: { batchSize: 2 } },
+      {
+        runDeleteMaintenanceTask: async () => {
+          deleteRunnerCalls += 1;
+        },
+      },
+    );
+    const internals = tracker as unknown as TrackerInternals;
+    seedTwoEntries(internals.db);
+    queueSubtitleLines(internals, 1);
+    let flushCalls = 0;
+    internals.flushNow = () => {
+      flushCalls += 1;
+      if (flushCalls > 1) throw new Error('bounded no-progress sentinel');
+    };
+
+    await assert.rejects(internals.deleteSession(1), /queue did not drain/i);
+
+    assert.equal(flushCalls, 1);
+    assert.equal(deleteRunnerCalls, 0);
+    assert.equal(internals.writeLock.locked, false);
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('reassignAnimeAnilist fails closed before resolving a conflict when writes cannot drain', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath, policy: { batchSize: 2 } });
+    const internals = tracker as unknown as TrackerInternals;
+    seedTwoEntries(internals.db);
+    internals.db.prepare('UPDATE imm_anime SET anilist_id = 123 WHERE anime_id = 2').run();
+    queueSubtitleLines(internals, 1);
+    internals.flushNow = () => {};
+
+    await assert.rejects(
+      internals.reassignAnimeAnilist(1, { anilistId: 123 }),
+      /queue did not drain/i,
+    );
+
+    assert.deepEqual(
+      internals.db
+        .prepare(
+          'SELECT anime_id AS animeId, anilist_id AS anilistId FROM imm_anime ORDER BY anime_id',
+        )
+        .all(),
+      [
+        { animeId: 1, anilistId: null },
+        { animeId: 2, anilistId: 123 },
+      ],
+    );
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('mergeAnime fails closed when queued writes cannot drain', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath, policy: { batchSize: 2 } });
+    const internals = tracker as unknown as TrackerInternals;
+    seedTwoEntries(internals.db);
+    queueSubtitleLines(internals, 1);
+    internals.flushNow = () => {};
+
+    await assert.rejects(internals.mergeAnime(1, [2]), /queue did not drain/i);
+
+    assert.deepEqual(
+      internals.db
+        .prepare('SELECT anime_id AS animeId FROM imm_anime ORDER BY anime_id')
+        .all()
+        .map((row) => (row as { animeId: number }).animeId),
+      [1, 2],
+    );
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('moveVideoToAnime fails closed when queued writes cannot drain', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath, policy: { batchSize: 2 } });
+    const internals = tracker as unknown as TrackerInternals;
+    seedTwoEntries(internals.db);
+    queueSubtitleLines(internals, 1);
+    internals.flushNow = () => {};
+
+    await assert.rejects(internals.moveVideoToAnime(2, 1), /queue did not drain/i);
+    assert.equal(
+      (
+        internals.db
+          .prepare('SELECT anime_id AS animeId FROM imm_videos WHERE video_id = 2')
+          .get() as {
+          animeId: number;
+        }
+      ).animeId,
+      2,
+    );
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('rebuildLifetimeSummaries fails closed when queued writes cannot drain', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath, policy: { batchSize: 2 } });
+    const internals = tracker as unknown as TrackerInternals;
+    seedTwoEntries(internals.db);
+    queueSubtitleLines(internals, 1);
+    internals.flushNow = () => {};
+
+    await assert.rejects(internals.rebuildLifetimeSummaries(), /queue did not drain/i);
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+function seedTwoEntries(db: DatabaseSync): void {
+  db.exec(`
+    INSERT INTO imm_anime (anime_id, normalized_title_key, canonical_title, CREATED_DATE, LAST_UPDATE_DATE)
+      VALUES (1, 'show', 'Show', 1000, 1000), (2, 'show season 1', 'Show Season 1', 1000, 1000);
+    INSERT INTO imm_videos (video_id, video_key, canonical_title, anime_id, source_type, watched, duration_ms, CREATED_DATE, LAST_UPDATE_DATE)
+      VALUES (1, 'local:/tmp/a.mkv', 'A', 1, 1, 0, 1440000, 1000, 1000),
+             (2, 'local:/tmp/b.mkv', 'B', 2, 1, 0, 1440000, 1000, 1000);
+    INSERT INTO imm_sessions (session_id, session_uuid, video_id, started_at_ms, ended_at_ms, status, active_watched_ms, CREATED_DATE, LAST_UPDATE_DATE)
+      VALUES (1, 'drain-session', 2, '1000', '2000', 2, 1000, 1000, 2000);
+  `);
+}
+
+function queueSubtitleLines(tracker: TrackerInternals, count: number): void {
+  for (let index = 0; index < count; index += 1) {
+    tracker.recordWrite({
+      kind: 'subtitleLine',
+      sessionId: 1,
+      videoId: 2,
+      lineIndex: index,
+      segmentStartMs: index * 1000,
+      segmentEndMs: index * 1000 + 900,
+      text: `line ${index}`,
+      wordOccurrences: [],
+      kanjiOccurrences: [],
+      firstSeen: 1000,
+      lastSeen: 2000,
+    });
+  }
+}
+
+/**
+ * Queued last so it sits past the first batch. Lifetime `total_lines_seen`
+ * reads this counter, not a COUNT over imm_subtitle_lines, so the rebuilt
+ * summary only reflects the session once the queue is drained all the way.
+ */
+function queueTelemetry(tracker: TrackerInternals, linesSeen: number): void {
+  tracker.recordWrite({
+    kind: 'telemetry',
+    sessionId: 1,
+    sampleMs: 3000,
+    lastMediaMs: 3000,
+    totalWatchedMs: 4000,
+    activeWatchedMs: 3500,
+    linesSeen,
+    tokensSeen: linesSeen * 5,
+    cardsMined: 2,
+    lookupCount: 0,
+    lookupHits: 0,
+    yomitanLookupCount: 0,
+    pauseCount: 0,
+    pauseMs: 0,
+    seekForwardCount: 0,
+    seekBackwardCount: 0,
+    mediaBufferEvents: 0,
+  });
+}
+
+/** The queued telemetry sample only exists in the database once the queue drained fully. */
+function latestTelemetryLinesSeen(db: DatabaseSync, sessionId: number): number | null {
+  const row = db
+    .prepare(
+      `SELECT lines_seen AS linesSeen
+       FROM imm_session_telemetry
+       WHERE session_id = ?
+       ORDER BY sample_ms DESC, telemetry_id DESC
+       LIMIT 1`,
+    )
+    .get(sessionId) as { linesSeen: number } | undefined;
+  return row ? Number(row.linesSeen) : null;
+}
+
+function countLinesForAnime(db: DatabaseSync, animeId: number): number {
+  const row = db
+    .prepare('SELECT COUNT(*) AS total FROM imm_subtitle_lines WHERE anime_id = ?')
+    .get(animeId) as { total: number };
+  return Number(row.total);
+}
+
+/**
+ * Both entry points must see a settled database before changing episode
+ * ownership. A single flushNow() only writes one batch off the front of the
+ * queue, so anything past `batchSize` would still be unwritten when the merge
+ * repoints rows.
+ */
+test('mergeAnime drains a queue larger than one batch before repointing rows', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath, policy: { batchSize: 2 } });
+    const internals = tracker as unknown as TrackerInternals;
+
+    seedTwoEntries(internals.db);
+    queueSubtitleLines(internals, 8);
+    queueTelemetry(internals, 8);
+    assert.ok(internals.queue.length > 2, 'expected more queued writes than one batch');
+
+    await internals.mergeAnime(1, [2]);
+
+    assert.equal(internals.queue.length, 0);
+    // Every queued line landed, attributed to the surviving entry.
+    assert.equal(countLinesForAnime(internals.db, 1), 8);
+    assert.equal(latestTelemetryLinesSeen(internals.db, 1), 8);
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('moveVideoToAnime drains a queue larger than one batch before repointing rows', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath, policy: { batchSize: 2 } });
+    const internals = tracker as unknown as TrackerInternals;
+
+    seedTwoEntries(internals.db);
+    queueSubtitleLines(internals, 8);
+    queueTelemetry(internals, 8);
+
+    await internals.moveVideoToAnime(2, 1);
+
+    assert.equal(internals.queue.length, 0);
+    assert.equal(countLinesForAnime(internals.db, 1), 8);
+    assert.equal(latestTelemetryLinesSeen(internals.db, 1), 8);
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});

src/core/services/__tests__/stats-server.test.ts

@@ -1053,6 +1053,55 @@ describe('stats server API routes', () => {
     assert.equal(body[0].canonicalTitle, 'Little Witch Academia');
   });
 
+  it('GET /api/stats/anime/merge-recommendations returns pending duplicate pairs', async () => {
+    const app = createStatsApp(
+      createMockTracker({
+        getAnimeMergeRecommendations: async () => [{ recommendationId: 4, animeIds: [1, 2] }],
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/anime/merge-recommendations');
+
+    assert.equal(res.status, 200);
+    assert.deepEqual(await res.json(), {
+      recommendations: [{ recommendationId: 4, animeIds: [1, 2] }],
+    });
+  });
+
+  it('DELETE /api/stats/anime/merge-recommendations/:id dismisses a pending pair', async () => {
+    let dismissedId: number | null = null;
+    const app = createStatsApp(
+      createMockTracker({
+        dismissAnimeMergeRecommendation: async (recommendationId: number) => {
+          dismissedId = recommendationId;
+          return true;
+        },
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/anime/merge-recommendations/4', {
+      method: 'DELETE',
+    });
+
+    assert.equal(res.status, 200);
+    assert.equal(dismissedId, 4);
+    assert.deepEqual(await res.json(), { ok: true });
+  });
+
+  it('DELETE /api/stats/anime/merge-recommendations/:id reports missing recommendations', async () => {
+    const app = createStatsApp(
+      createMockTracker({
+        dismissAnimeMergeRecommendation: async () => false,
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/anime/merge-recommendations/99', {
+      method: 'DELETE',
+    });
+
+    assert.equal(res.status, 404);
+  });
+
   it('GET /api/stats/anime/:animeId returns anime detail with episodes', async () => {
     const app = createStatsApp(createMockTracker());
     const res = await app.request('/api/stats/anime/1');
@@ -2454,6 +2503,80 @@ Aligned English subtitle
     });
   });
 
+  it('POST /api/stats/mine-card marks the configured Kiku word card kind', async () => {
+    await withTempDir(async (dir) => {
+      const sourcePath = path.join(dir, 'episode.mkv');
+      fs.writeFileSync(sourcePath, 'fake media');
+
+      await withFakeAnkiConnect(
+        async (requests, url) => {
+          const app = createStatsApp(createMockTracker(), {
+            addYomitanNote: async () => 777,
+            createMediaGenerator: () => ({
+              generateAudio: async () => null,
+              generateScreenshot: async () => null,
+              generateAnimatedImage: async () => null,
+            }),
+            ankiConnectConfig: {
+              url,
+              deck: 'Mining',
+              fields: {
+                image: 'Picture',
+                sentence: 'Sentence',
+              },
+              media: {
+                generateAudio: false,
+                generateImage: false,
+              },
+              isKiku: {
+                enabled: true,
+                fieldGrouping: 'disabled',
+                deleteDuplicateInAuto: true,
+              },
+              lapisKiku: {
+                wordCardKind: 'click',
+              },
+            },
+          });
+
+          const res = await app.request('/api/stats/mine-card?mode=word', {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: JSON.stringify({
+              sourcePath,
+              startMs: 1_000,
+              endMs: 2_000,
+              sentence: '猫を見た',
+              word: '猫',
+              videoTitle: 'Episode 1',
+            }),
+          });
+
+          const body = await res.json();
+          assert.equal(res.status, 200, JSON.stringify(body));
+
+          const updateRequest = requests.find((request) => request.action === 'updateNoteFields');
+          const fields = updateRequest?.params?.note?.fields ?? {};
+          assert.equal(fields.IsClickCard, 'x');
+          assert.equal(fields.IsWordAndSentenceCard, '');
+          assert.equal(fields.IsSentenceCard, '');
+          assert.equal(fields.IsAudioCard, '');
+        },
+        {
+          notesInfoFields: {
+            Expression: { value: '猫' },
+            Sentence: { value: '' },
+            Picture: { value: '' },
+            IsWordAndSentenceCard: { value: '' },
+            IsClickCard: { value: '' },
+            IsSentenceCard: { value: '' },
+            IsAudioCard: { value: '' },
+          },
+        },
+      );
+    });
+  });
+
   it('POST /api/stats/mine-card writes word mining sentence audio and image together', async () => {
     await withTempDir(async (dir) => {
       const sourcePath = path.join(dir, 'episode.mkv');
@@ -2950,6 +3073,148 @@ Aligned English subtitle
     assert.equal(deleteCalls, 0);
   });
 
+  it('POST /api/stats/anime/:animeId/merge folds the given entries into the target', async () => {
+    let merged: { targetAnimeId: number; sourceAnimeIds: number[] } | null = null;
+    const app = createStatsApp(
+      createMockTracker({
+        mergeAnime: async (targetAnimeId: number, sourceAnimeIds: number[]) => {
+          merged = { targetAnimeId, sourceAnimeIds };
+          return {
+            survivingAnimeId: targetAnimeId,
+            mergedAnimeIds: sourceAnimeIds,
+            movedVideos: 3,
+          };
+        },
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/anime/7/merge', {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      // The target repeated in the sources must not delete the entry we keep.
+      body: '{"sourceAnimeIds":[8,9,8,7]}',
+    });
+
+    assert.equal(res.status, 200);
+    assert.deepEqual(merged, { targetAnimeId: 7, sourceAnimeIds: [8, 9] });
+    assert.deepEqual(await res.json(), {
+      ok: true,
+      animeId: 7,
+      mergedAnimeIds: [8, 9],
+      movedVideos: 3,
+    });
+  });
+
+  it('POST /api/stats/anime/:animeId/merge rejects an empty or malformed source list', async () => {
+    let mergeCalls = 0;
+    const app = createStatsApp(
+      createMockTracker({
+        mergeAnime: async () => {
+          mergeCalls += 1;
+          return { survivingAnimeId: 7, mergedAnimeIds: [], movedVideos: 0 };
+        },
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    for (const body of [
+      '{"sourceAnimeIds":[]}',
+      '{"sourceAnimeIds":[7]}',
+      '{"sourceAnimeIds":0}',
+    ]) {
+      const res = await app.request('/api/stats/anime/7/merge', {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body,
+      });
+      assert.equal(res.status, 400);
+    }
+    assert.equal(mergeCalls, 0);
+  });
+
+  it('PATCH /api/stats/media/:videoId/anime moves the episode to another entry', async () => {
+    let moved: { videoId: number; animeId: number } | null = null;
+    const app = createStatsApp(
+      createMockTracker({
+        moveVideoToAnime: async (videoId: number, animeId: number) => {
+          moved = { videoId, animeId };
+          return { targetAnimeId: animeId, previousAnimeId: 4, removedPreviousAnime: true };
+        },
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/media/12/anime', {
+      method: 'PATCH',
+      headers: { 'Content-Type': 'application/json' },
+      body: '{"animeId":7}',
+    });
+
+    assert.equal(res.status, 200);
+    assert.deepEqual(moved, { videoId: 12, animeId: 7 });
+    assert.deepEqual(await res.json(), {
+      ok: true,
+      animeId: 7,
+      previousAnimeId: 4,
+      removedPreviousAnime: true,
+    });
+  });
+
+  it('POST /api/stats/anime/:animeId/merge reports a merge that folded nothing as 404', async () => {
+    const app = createStatsApp(
+      createMockTracker({
+        mergeAnime: async (targetAnimeId: number) => ({
+          survivingAnimeId: targetAnimeId,
+          mergedAnimeIds: [],
+          movedVideos: 0,
+        }),
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/anime/7/merge', {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: '{"sourceAnimeIds":[8]}',
+    });
+
+    assert.equal(res.status, 404);
+  });
+
+  it('PATCH /api/stats/media/:videoId/anime reports an unknown target as 404', async () => {
+    const app = createStatsApp(
+      createMockTracker({
+        moveVideoToAnime: async () => {
+          throw new Error('Unknown episode or target library entry');
+        },
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/media/12/anime', {
+      method: 'PATCH',
+      headers: { 'Content-Type': 'application/json' },
+      body: '{"animeId":99}',
+    });
+
+    assert.equal(res.status, 404);
+  });
+
+  it('PATCH /api/stats/media/:videoId/anime does not disguise storage failures as 404', async () => {
+    const app = createStatsApp(
+      createMockTracker({
+        moveVideoToAnime: async () => {
+          throw new Error('database is locked');
+        },
+      } as Partial<ImmersionTrackerService>),
+    );
+
+    const res = await app.request('/api/stats/media/12/anime', {
+      method: 'PATCH',
+      headers: { 'Content-Type': 'application/json' },
+      body: '{"animeId":7}',
+    });
+
+    assert.notEqual(res.status, 404);
+    assert.equal(res.status >= 500, true);
+  });
+
   it('POST /api/stats/anki/browse returns 400 for missing noteId', async () => {
     const app = createStatsApp(createMockTracker());
     const res = await app.request('/api/stats/anki/browse', { method: 'POST' });

src/core/services/anilist/cover-art-fetcher.ts

@@ -327,6 +327,7 @@ export function createCoverArtFetcher(
         titleEnglish: selected.title?.english ?? null,
         titleNative: selected.title?.native ?? null,
         episodesTotal: selected.episodes ?? null,
+        exactTitleMatch: resolution?.exactTitleMatch ?? false,
       });
 
       logger.info(

src/core/services/anilist/season-resolver.test.ts

@@ -156,6 +156,79 @@ test('season 1 resolves to the anchor without relation lookups', async () => {
   assert.deepEqual(relationLookups, []);
 });
 
+test('a sequel resolution is not certified by the anchor exact-title evidence', async () => {
+  // The anchor matched the search title exactly, but the hopped-to entry is a
+  // different inference (a split-cour chain can land one season short), so the
+  // sequel result must report its own title evidence, not the anchor's.
+  const { execute } = createExecutor(OREGAIRU_SEARCH, OREGAIRU_RELATIONS);
+  const result = await resolveAnilistSeasonMedia(
+    { title: 'My Teen Romantic Comedy SNAFU', season: 2, episode: 1 },
+    { execute },
+  );
+
+  assert.equal(result?.id, 20698);
+  assert.equal(result?.via, 'sequel-chain');
+  assert.equal(result?.exactTitleMatch, false);
+});
+
+test('a sequel resolution whose own title matches the parsed title stays exact', async () => {
+  const anchor: AnilistSeasonMedia = {
+    id: 1,
+    episodes: 12,
+    format: 'TV',
+    title: { english: 'Show' },
+  };
+  const sequel: AnilistSeasonMedia = {
+    id: 2,
+    episodes: 12,
+    format: 'TV',
+    title: { english: 'Show 2nd Season' },
+  };
+  const { execute } = createExecutor([anchor], {
+    1: [{ relationType: 'SEQUEL', node: sequel }],
+  });
+
+  const result = await resolveAnilistSeasonMedia(
+    { title: 'Show 2nd Season', season: 2, episode: 1 },
+    { execute },
+  );
+
+  assert.equal(result?.id, 2);
+  assert.equal(result?.via, 'sequel-chain');
+  assert.equal(result?.exactTitleMatch, true);
+});
+
+test('reports an exact normalized synonym match as strong evidence', async () => {
+  const { execute } = createExecutor([
+    {
+      id: 1,
+      episodes: 12,
+      format: 'TV',
+      title: { english: 'Hitori Gotoh Story' },
+      synonyms: ['BOCCHI THE ROCK'],
+    },
+  ]);
+
+  const result = await resolveAnilistSeasonMedia({ title: 'Bocchi the Rock!' }, { execute });
+
+  assert.equal(result?.exactTitleMatch, true);
+});
+
+test('reports a fuzzy-only search result as weak evidence', async () => {
+  const { execute } = createExecutor([
+    {
+      id: 1,
+      episodes: 12,
+      format: 'TV',
+      title: { english: 'Actual Show' },
+    },
+  ]);
+
+  const result = await resolveAnilistSeasonMedia({ title: 'Unrelated Release' }, { execute });
+
+  assert.equal(result?.exactTitleMatch, false);
+});
+
 test('strips a season marker already present in the parsed title', async () => {
   const { execute, searches } = createExecutor(OREGAIRU_SEARCH, OREGAIRU_RELATIONS);
   const result = await resolveAnilistSeasonMedia(

src/core/services/anilist/season-resolver.ts

@@ -9,6 +9,8 @@
  * reports `seasonResolved: false` so callers can refuse to act instead of guessing.
  */
 
+import { normalizeTitleIdentity } from '../../utils/title-normalization';
+
 export interface AnilistSeasonMediaTitle {
   romaji?: string | null;
   english?: string | null;
@@ -42,6 +44,8 @@ export interface AnilistSeasonResolution {
   seasonResolved: boolean;
   requestedSeason: number | null;
   via: AnilistSeasonResolutionVia;
+  /** Exact normalized match against an AniList title or synonym. */
+  exactTitleMatch: boolean;
 }
 
 export interface ResolveAnilistSeasonMediaInput {
@@ -115,10 +119,6 @@ const SEASONAL_FORMAT_PRIORITY = ['TV', 'TV_SHORT', 'ONA'];
 
 const MAX_SEQUEL_HOPS = 12;
 
-function normalizeTitle(value: string): string {
-  return value.trim().toLowerCase().replace(/\s+/g, ' ');
-}
-
 /**
  * Drops season markers a release name carries but AniList titles never do,
  * so "Some Show Season 3" and "Some Show S3" both search as "Some Show".
@@ -136,7 +136,7 @@ function mediaTitles(media: AnilistSeasonMedia): string[] {
   const synonyms = Array.isArray(media.synonyms) ? media.synonyms : [];
   return [media.title?.english, media.title?.romaji, media.title?.native, ...synonyms]
     .filter((value): value is string => typeof value === 'string' && value.trim().length > 0)
-    .map((value) => normalizeTitle(value));
+    .map((value) => normalizeTitleIdentity(value));
 }
 
 function displayTitle(media: AnilistSeasonMedia, fallback: string): string {
@@ -176,6 +176,7 @@ function toResolution(
   season: number | null,
   via: AnilistSeasonResolutionVia,
   seasonResolved: boolean,
+  exactTitleMatch: boolean,
 ): AnilistSeasonResolution {
   return {
     id: media.id,
@@ -185,6 +186,7 @@ function toResolution(
     seasonResolved,
     requestedSeason: season,
     via,
+    exactTitleMatch,
   };
 }
 
@@ -209,9 +211,10 @@ export function pickAnchorMedia(
       : media;
   const pool = episodeFiltered.length > 0 ? episodeFiltered : media;
 
-  const targets = [normalizeTitle(title), normalizeTitle(stripSeasonSuffix(title))].filter(
+  const targets = [
-    (value, index, all) => value.length > 0 && all.indexOf(value) === index,
+    normalizeTitleIdentity(title),
-  );
+    normalizeTitleIdentity(stripSeasonSuffix(title)),
+  ].filter((value, index, all) => value.length > 0 && all.indexOf(value) === index);
 
   const scored = pool.map((entry, index) => {
     const candidateTitles = mediaTitles(entry);
@@ -367,9 +370,20 @@ export async function resolveAnilistSeasonMedia(
     episode: season === null || season <= 1 ? input.episode : null,
   });
   if (!anchor) return null;
+  // Certifies the media actually returned, never the anchor on its behalf: a
+  // sequel-chain hop can land one season short (split-cour entries) while the
+  // anchor title still matches perfectly, and that certainty must not carry
+  // over to the hopped-to entry.
+  const exactMatchFor = (candidate: AnilistSeasonMedia): boolean => {
+    const titles = mediaTitles(candidate);
+    return (
+      titles.includes(normalizeTitleIdentity(searchTitle)) ||
+      titles.includes(normalizeTitleIdentity(input.title))
+    );
+  };
 
   if (season === null || season <= 1) {
-    return toResolution(anchor, searchTitle, season, 'anchor', true);
+    return toResolution(anchor, searchTitle, season, 'anchor', true, exactMatchFor(anchor));
   }
 
   let chainError: unknown = null;
@@ -383,7 +397,14 @@ export async function resolveAnilistSeasonMedia(
     deps.logInfo?.(
       `[anilist] season ${season} of "${searchTitle}" resolved via sequel chain: ${displayTitle(viaChain, searchTitle)} (${viaChain.id})`,
     );
-    return toResolution(viaChain, searchTitle, season, 'sequel-chain', true);
+    return toResolution(
+      viaChain,
+      searchTitle,
+      season,
+      'sequel-chain',
+      true,
+      exactMatchFor(viaChain),
+    );
   }
 
   const viaAirOrder = pickByAirOrder(anchor, season, media);
@@ -391,7 +412,14 @@ export async function resolveAnilistSeasonMedia(
     deps.logInfo?.(
       `[anilist] season ${season} of "${searchTitle}" resolved via air order: ${displayTitle(viaAirOrder, searchTitle)} (${viaAirOrder.id})`,
     );
-    return toResolution(viaAirOrder, searchTitle, season, 'air-order', true);
+    return toResolution(
+      viaAirOrder,
+      searchTitle,
+      season,
+      'air-order',
+      true,
+      exactMatchFor(viaAirOrder),
+    );
   }
 
   // The chain failed for transport reasons rather than because the season is absent;
@@ -403,5 +431,5 @@ export async function resolveAnilistSeasonMedia(
   deps.logInfo?.(
     `[anilist] could not resolve season ${season} of "${searchTitle}"; falling back to ${displayTitle(anchor, searchTitle)} (${anchor.id})`,
   );
-  return toResolution(anchor, searchTitle, season, 'anchor', false);
+  return toResolution(anchor, searchTitle, season, 'anchor', false, exactMatchFor(anchor));
 }

src/core/services/ass-text.test.ts

@@ -0,0 +1,195 @@
+import { test } from 'node:test';
+import assert from 'node:assert/strict';
+import {
+  assOverrideSignature,
+  assToPlainText,
+  collectAssOverrideCommands,
+  extractAssOverrideBlocks,
+  hasAssTemporalOverride,
+  isAnimatedAssEffectKind,
+  isAssTemporalCommand,
+  normalizePlainSubtitleText,
+  parseAssEffectField,
+} from './ass-text';
+
+test('assToPlainText drops vector drawing runs', () => {
+  assert.equal(
+    assToPlainText(
+      '{\\an5\\pos(730,1042)\\p1\\blur1}m 20 0 b 10 0 0 10 0 20 b 0 31 10 40 20 40 {\\p0}',
+    ),
+    '',
+  );
+});
+
+test('assToPlainText keeps text around drawing runs on the same event', () => {
+  assert.equal(
+    assToPlainText('{\\p1}m 0 0 l 10 10{\\p0}本文{\\p1}m 5 5 l 6 6{\\p0}続き'),
+    '本文続き',
+  );
+});
+
+test('assToPlainText leaves \\pos alone when no drawing mode is active', () => {
+  assert.equal(assToPlainText('{\\pos(960,1068)\\bord3}位置指定'), '位置指定');
+});
+
+test('assToPlainText does not read \\pos as a drawing tag', () => {
+  assert.equal(assToPlainText('{\\p1\\pos(1,2)}m 0 0 l 5 5'), '');
+});
+
+test('assToPlainText resolves line-break and space escapes', () => {
+  assert.equal(assToPlainText('一行目\\N二行目'), '一行目\n二行目');
+  assert.equal(assToPlainText('一行目\\n二行目'), '一行目\n二行目');
+  assert.equal(assToPlainText('一行目\\N二行目', ' '), '一行目 二行目');
+  assert.equal(assToPlainText('間\\h隔'), '間 隔');
+});
+
+test('assToPlainText matches mpv on brace and backslash sequences', () => {
+  // mpv has no `\{` / `\}` / `\\` escapes: the backslashes are literal text and the
+  // braces still open and close an override block.
+  assert.equal(assToPlainText('\\{注\\}'), '\\');
+  assert.equal(assToPlainText('\\\\N'), '\\\n');
+});
+
+test('assToPlainText renders an unclosed override block verbatim', () => {
+  // mpv shows the stray brace; guessing where the block ended can eat a whole line.
+  assert.equal(assToPlainText('本文{\\pos(1,2)'), '本文{\\pos(1,2)');
+});
+
+test('assToPlainText is idempotent', () => {
+  const samples = [
+    '{\\an5\\p1}m 0 0 l 5 5{\\p0}本文',
+    '\\{注\\}',
+    '\\\\N',
+    '本文{\\pos(1,2)',
+    '一行目\\N二行目\\h終わり',
+  ];
+
+  for (const sample of samples) {
+    const once = assToPlainText(sample);
+    assert.equal(assToPlainText(once), once, sample);
+  }
+});
+
+test('assToPlainText normalizes CRLF before converting', () => {
+  assert.equal(assToPlainText('一行目\r\n二行目'), '一行目\n二行目');
+});
+
+test('normalizePlainSubtitleText settles whitespace without decoding ASS', () => {
+  // A brace reaching this layer is literal text mpv chose to show, not markup.
+  assert.equal(normalizePlainSubtitleText('本文{\\pos(1,2)'), '本文{\\pos(1,2)');
+  assert.equal(normalizePlainSubtitleText('一行目\\N二行目'), '一行目\n二行目');
+  assert.equal(
+    normalizePlainSubtitleText('一行目\\N二行目', { collapseLineBreaks: true }),
+    '一行目 二行目',
+  );
+  assert.equal(normalizePlainSubtitleText('  余白  ', { trim: false }), '  余白  ');
+});
+
+test('normalizePlainSubtitleText is idempotent', () => {
+  for (const sample of ['一行目\\N二行目', '間\\h隔', '本文{\\pos(1,2)', '  余白  ']) {
+    const once = normalizePlainSubtitleText(sample);
+    assert.equal(normalizePlainSubtitleText(once), once, sample);
+  }
+});
+
+test('extractAssOverrideBlocks returns block contents', () => {
+  assert.deepEqual(extractAssOverrideBlocks('{\\an8}上{\\fad(200,200)}下'), [
+    '\\an8',
+    '\\fad(200,200)',
+  ]);
+  assert.deepEqual(extractAssOverrideBlocks('括弧なし'), []);
+});
+
+test('collectAssOverrideCommands captures names and arguments from blocks only', () => {
+  const commands = collectAssOverrideCommands('{\\pos(1,2)\\1c&HFFFFFF&\\kf30}歌詞');
+
+  assert.deepEqual(commands, [
+    { name: 'pos', args: '1,2', animated: false },
+    { name: '1c', args: '&HFFFFFF&', animated: false },
+    { name: 'kf', args: '30', animated: false },
+  ]);
+
+  // A `\pos(...)` sitting in visible text is not typesetting markup.
+  assert.deepEqual(collectAssOverrideCommands('\\pos(730,1042) と書いてある'), []);
+});
+
+test('collectAssOverrideCommands marks tags animated by a wrapping \\t', () => {
+  const commands = collectAssOverrideCommands('{\\clip(0,0,10,10)\\t(0,500,\\frz30)}文字');
+
+  assert.deepEqual(
+    commands.map((command) => [command.name, command.animated]),
+    [
+      ['clip', false],
+      ['t', false],
+      ['frz', true],
+    ],
+  );
+  assert.equal(hasAssTemporalOverride(commands), true);
+});
+
+test('collectAssOverrideCommands stops descending into deeply nested \\t tags', () => {
+  // Nested far past the recursion cap. Uncapped, this recurses once per level, and a
+  // pathological line (real files reach one or two levels) overflows the stack.
+  const nesting = 32;
+  const block = `{${'\\t(0,500,'.repeat(nesting)}\\frz30${')'.repeat(nesting)}}文字`;
+
+  const commands = collectAssOverrideCommands(block);
+
+  // The outer `\t` plus one per allowed recursion level, and nothing from below the cap.
+  assert.equal(commands.length, 9);
+  assert.deepEqual(new Set(commands.map((command) => command.name)), new Set(['t']));
+  assert.equal(hasAssTemporalOverride(commands), true);
+});
+
+test('hasAssTemporalOverride ignores static placement and shape tags', () => {
+  assert.equal(
+    hasAssTemporalOverride(collectAssOverrideCommands('{\\pos(1,2)\\clip(m 1 1)\\blur2}文字')),
+    false,
+  );
+  assert.equal(hasAssTemporalOverride(collectAssOverrideCommands('{\\move(1,2,3,4)}文字')), true);
+});
+
+test('isAssTemporalCommand covers only intrinsically animated tags', () => {
+  for (const command of ['t', 'move', 'k', 'kf', 'ko', 'K']) {
+    assert.equal(isAssTemporalCommand(command), true, command);
+  }
+  for (const command of ['clip', 'iclip', 'frz', 'fscx', 'blur', 'be', 'pos', 'fad']) {
+    assert.equal(isAssTemporalCommand(command), false, command);
+  }
+});
+
+test('assOverrideSignature distinguishes events by their override values', () => {
+  const first = assOverrideSignature(collectAssOverrideCommands('{\\clip(m 1 1)}歌詞'));
+  const second = assOverrideSignature(collectAssOverrideCommands('{\\clip(m 2 2)}歌詞'));
+  const repeat = assOverrideSignature(collectAssOverrideCommands('{\\clip(m 1 1)}別の行'));
+
+  assert.notEqual(first, second);
+  assert.equal(first, repeat);
+});
+
+test('parseAssEffectField classifies the event-level Effect column', () => {
+  assert.equal(parseAssEffectField(''), 'none');
+  assert.equal(parseAssEffectField('   '), 'none');
+  assert.equal(parseAssEffectField('Banner;20;1;0'), 'banner');
+  assert.equal(parseAssEffectField('Scroll up;0;0;30;10'), 'scroll');
+  assert.equal(parseAssEffectField('Scroll down;0;0;30;10'), 'scroll');
+  assert.equal(parseAssEffectField('Karaoke'), 'karaoke');
+  assert.equal(parseAssEffectField('fx-template'), 'other');
+});
+
+test('parseAssEffectField matches stock effect names exactly', () => {
+  // Custom effect names that merely start with a stock name are not stock effects.
+  assert.equal(parseAssEffectField('scrolling-credit'), 'other');
+  assert.equal(parseAssEffectField('bannerfx;1'), 'other');
+  assert.equal(parseAssEffectField('karaoke-template'), 'other');
+  assert.equal(parseAssEffectField('Scroll'), 'other');
+});
+
+test('isAnimatedAssEffectKind covers the stock animated effects only', () => {
+  assert.equal(isAnimatedAssEffectKind('karaoke'), true);
+  assert.equal(isAnimatedAssEffectKind('banner'), true);
+  assert.equal(isAnimatedAssEffectKind('scroll'), true);
+  // Typesetting groups put static template names in this column too.
+  assert.equal(isAnimatedAssEffectKind('other'), false);
+  assert.equal(isAnimatedAssEffectKind('none'), false);
+});

src/core/services/ass-text.ts

@@ -0,0 +1,280 @@
+/*
+ * ASS/SSA text handling, split into two deliberately distinct contracts:
+ *
+ *   assToPlainText()             raw ASS event text -> plain text. Ingestion only.
+ *   normalizePlainSubtitleText() already-decoded text -> display/lookup form.
+ *
+ * Subtitle text is decoded from ASS exactly once, at the point it enters the app: the
+ * file cue parser does it for sidecar/embedded scripts, and mpv does it for live text
+ * (`sub-text` is already run through mpv's own `ass_to_plaintext`). Everything
+ * downstream -- renderer, timing tracker, tokenizer, tokenization cache keys -- gets
+ * plain text and only normalizes whitespace, so no layer decodes the same string twice.
+ *
+ * assToPlainText mirrors mpv's `ass_to_plaintext` rather than inventing its own rules,
+ * so a cue parsed from a file reads the same as the same line arriving live:
+ *   - `{...}` override blocks are markup
+ *   - `\pN ... \p0` runs are vector paths, not dialogue
+ *   - `\N`, `\n` and `\h` are the only escapes; `\{`, `\}` and `\\` are NOT escapes,
+ *     so `\{注\}` decodes to a lone backslash exactly as mpv renders it
+ *   - an unclosed `{` is rendered verbatim instead of swallowing the rest of the line
+ * Because the decoder never emits an escape or a closed brace, running it twice is a
+ * no-op -- but downstream code should still use normalizePlainSubtitleText.
+ */
+
+/** What `\N` and `\n` become. */
+export type AssLineBreak = '\n' | ' ';
+
+// `\p<n>` with n > 0 switches libass into vector-drawing mode: everything until the
+// next `\p0` is a path (`m 20 0 b 10 0 ...`), not dialogue. The negative lookahead keeps
+// `\pos(...)` from being read as a drawing tag.
+const ASS_DRAWING_SCALE_PATTERN = /\\p(?![a-zA-Z])(\d*)/g;
+
+function readDrawingScale(block: string): number | null {
+  ASS_DRAWING_SCALE_PATTERN.lastIndex = 0;
+  let scale: number | null = null;
+  let match: RegExpExecArray | null;
+  // Drawing mode is whatever the last `\p` tag in this block set it to.
+  while ((match = ASS_DRAWING_SCALE_PATTERN.exec(block)) !== null) {
+    scale = match[1] ? Number(match[1]) : 0;
+  }
+  return scale;
+}
+
+/** Resolve `\N`, `\n` and `\h`. The only text-level escapes libass recognises. */
+function resolveWhitespaceEscapes(text: string, lineBreak: AssLineBreak): string {
+  return text.replace(/\\([Nnh])/g, (_match, escaped: string) =>
+    escaped === 'h' ? ' ' : lineBreak,
+  );
+}
+
+/** Strip `{...}` override blocks and the drawing runs they enable. */
+function stripAssMarkup(raw: string): string {
+  let out = '';
+  let cursor = 0;
+  let drawing = false;
+
+  while (cursor < raw.length) {
+    if (raw[cursor] !== '{') {
+      if (!drawing) {
+        out += raw[cursor];
+      }
+      cursor += 1;
+      continue;
+    }
+
+    const close = raw.indexOf('}', cursor + 1);
+    if (close === -1) {
+      // mpv shows an unclosed `{` and everything after it. Guessing where the block was
+      // meant to end can eat a whole line of dialogue.
+      if (!drawing) {
+        out += raw.slice(cursor);
+      }
+      break;
+    }
+
+    const scale = readDrawingScale(raw.slice(cursor, close + 1));
+    if (scale !== null) {
+      drawing = scale > 0;
+    }
+    cursor = close + 1;
+  }
+
+  return out;
+}
+
+/**
+ * Decode a raw ASS/SSA event text field. Call this once, where the text enters the app;
+ * downstream layers take the result as plain text.
+ */
+export function assToPlainText(text: string, lineBreak: AssLineBreak = '\n'): string {
+  if (!text) return '';
+  return resolveWhitespaceEscapes(stripAssMarkup(text.replace(/\r\n/g, '\n')), lineBreak);
+}
+
+export interface NormalizePlainSubtitleTextOptions {
+  /** Fold every line break into a single space. */
+  collapseLineBreaks?: boolean;
+  trim?: boolean;
+}
+
+/**
+ * Whitespace normalization for text that has already been decoded -- by mpv for live
+ * subtitles, by the cue parser for files. Override blocks and drawing runs are none of
+ * this function's business; a `{` that reaches here is literal text mpv chose to show.
+ *
+ * `\N`/`\n`/`\h` are still folded, because subtitle sources outside the ASS path (asbplayer
+ * and other websocket clients) forward them raw and the display layer has to cope.
+ */
+export function normalizePlainSubtitleText(
+  text: string,
+  options: NormalizePlainSubtitleTextOptions = {},
+): string {
+  if (!text) return '';
+  const { collapseLineBreaks = false, trim = true } = options;
+
+  let normalized = resolveWhitespaceEscapes(
+    text.replace(/\r\n/g, '\n'),
+    collapseLineBreaks ? ' ' : '\n',
+  );
+  if (collapseLineBreaks) {
+    normalized = normalized.replace(/\n/g, ' ').replace(/\s+/g, ' ');
+  }
+
+  return trim ? normalized.trim() : normalized;
+}
+
+/** The contents of each `{...}` block, without the braces. */
+export function extractAssOverrideBlocks(text: string): string[] {
+  const blocks: string[] = [];
+  let cursor = 0;
+
+  while (cursor < text.length) {
+    const open = text.indexOf('{', cursor);
+    if (open === -1) {
+      break;
+    }
+    const close = text.indexOf('}', open + 1);
+    if (close === -1) {
+      break;
+    }
+    blocks.push(text.slice(open + 1, close));
+    cursor = close + 1;
+  }
+
+  return blocks;
+}
+
+export interface AssOverrideCommand {
+  /** Tag name without the backslash, e.g. `pos`, `kf`, `1c`. */
+  name: string;
+  /** Everything the tag was given, e.g. `960,1068` for `\pos(960,1068)`. */
+  args: string;
+  /** Nested inside a `\t(...)` argument, so its value is animated over the event. */
+  animated: boolean;
+}
+
+const ASS_OVERRIDE_NAME_PATTERN = /[1-4]?[a-zA-Z]+/y;
+
+function readCommandArgs(block: string, start: number): { args: string; next: number } {
+  if (block[start] === '(') {
+    let depth = 0;
+    for (let i = start; i < block.length; i += 1) {
+      if (block[i] === '(') depth += 1;
+      else if (block[i] === ')') {
+        depth -= 1;
+        if (depth === 0) {
+          return { args: block.slice(start + 1, i), next: i + 1 };
+        }
+      }
+    }
+    return { args: block.slice(start + 1), next: block.length };
+  }
+
+  const nextTag = block.indexOf('\\', start);
+  const end = nextTag === -1 ? block.length : nextTag;
+  return { args: block.slice(start, end), next: end };
+}
+
+// `\t(...)` can wrap another `\t(...)`, and nothing in the format stops an author (or a
+// malformed file) from nesting them thousands deep. Real typesetting never goes past one
+// or two levels, so stop recursing well before the call stack is at risk.
+const MAX_ANIMATION_NESTING_DEPTH = 8;
+
+function parseOverrideBlock(
+  block: string,
+  animated: boolean,
+  into: AssOverrideCommand[],
+  depth = 0,
+): void {
+  let cursor = 0;
+
+  while (cursor < block.length) {
+    if (block[cursor] !== '\\') {
+      cursor += 1;
+      continue;
+    }
+
+    ASS_OVERRIDE_NAME_PATTERN.lastIndex = cursor + 1;
+    const nameMatch = ASS_OVERRIDE_NAME_PATTERN.exec(block);
+    if (!nameMatch) {
+      cursor += 1;
+      continue;
+    }
+
+    const name = nameMatch[0];
+    const { args, next } = readCommandArgs(block, cursor + 1 + name.length);
+    into.push({ name, args: args.trim(), animated });
+    // `\t(0,500,\frz30)` animates whatever it wraps, so record the inner tags too.
+    if (name === 't' && args.includes('\\') && depth < MAX_ANIMATION_NESTING_DEPTH) {
+      parseOverrideBlock(args, true, into, depth + 1);
+    }
+    cursor = next;
+  }
+}
+
+/**
+ * Override commands with their arguments, in source order. Only `{...}` blocks are
+ * inspected, so a `\pos(...)` sitting in visible text is never mistaken for markup.
+ */
+export function collectAssOverrideCommands(text: string): AssOverrideCommand[] {
+  const commands: AssOverrideCommand[] = [];
+  for (const block of extractAssOverrideBlocks(text)) {
+    parseOverrideBlock(block, false, commands);
+  }
+  return commands;
+}
+
+// Tags that are animated by definition: `\t` interpolates, `\move` travels, and the
+// karaoke tags advance a highlight across the event's own duration. Everything else --
+// `\pos`, `\clip`, `\frz`, `\blur`, `\fad` -- is a static value for the event, so its
+// presence says nothing about whether neighbouring events form one animation.
+const ASS_TEMPORAL_COMMANDS = new Set(['t', 'move', 'k', 'kf', 'ko', 'K']);
+
+export function isAssTemporalCommand(name: string): boolean {
+  return ASS_TEMPORAL_COMMANDS.has(name);
+}
+
+/** True when the event animates on its own, or animates a static tag through `\t(...)`. */
+export function hasAssTemporalOverride(commands: readonly AssOverrideCommand[]): boolean {
+  return commands.some((command) => command.animated || isAssTemporalCommand(command.name));
+}
+
+/**
+ * Canonical form of an event's override values, for comparing consecutive events. Two
+ * events with the same signature were typeset identically, so neither is a frame of an
+ * animation the other belongs to.
+ */
+export function assOverrideSignature(commands: readonly AssOverrideCommand[]): string {
+  return commands.map((command) => `${command.name}(${command.args})`).join('|');
+}
+
+export type AssEffectKind = 'none' | 'banner' | 'scroll' | 'karaoke' | 'other';
+
+// The stock effects, matched exactly. Typesetting groups put their own template names in
+// this column -- `scrolling-credit` is a static sign, not libass's `Scroll up` -- so a
+// prefix match would hand out animation evidence to arbitrary custom effects.
+const STOCK_ASS_EFFECTS = new Map<string, AssEffectKind>([
+  ['banner', 'banner'],
+  ['scroll up', 'scroll'],
+  ['scroll down', 'scroll'],
+  ['karaoke', 'karaoke'],
+]);
+
+/**
+ * The event-level `Effect` column. The stock values (`Banner;...`, `Scroll up;...`,
+ * `Scroll down;...`, `Karaoke`) all animate; anything else is a custom name and lands in
+ * `other`.
+ */
+export function parseAssEffectField(raw: string): AssEffectKind {
+  const value = raw.trim().toLowerCase();
+  if (!value) return 'none';
+
+  const name = value.split(';', 1)[0]!.trim();
+  return STOCK_ASS_EFFECTS.get(name) ?? 'other';
+}
+
+const ANIMATED_ASS_EFFECT_KINDS = new Set<AssEffectKind>(['banner', 'scroll', 'karaoke']);
+
+export function isAnimatedAssEffectKind(kind: AssEffectKind): boolean {
+  return ANIMATED_ASS_EFFECT_KINDS.has(kind);
+}

src/core/services/config-hot-reload.ts

@@ -85,6 +85,7 @@ const HOT_RELOAD_EXACT_OR_PREFIX_PATHS = [
   'ankiConnect.fields.miscInfo',
   'ankiConnect.isLapis.sentenceCardModel',
   'ankiConnect.isKiku.fieldGrouping',
+  'ankiConnect.lapisKiku.wordCardKind',
 ] as const;
 
 function hotReloadFieldForChangedPath(path: string): string | null {

src/core/services/immersion-tracker-service.test.ts

@@ -1414,6 +1414,353 @@ test('deleteSession ignores the currently active session and keeps new writes fl
   }
 });
 
+test('deleteSession yields the main event loop while delete maintenance is pending', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  const deleteGate: { release?: () => void } = {};
+  let deleteRunnerCalled = false;
+  let bufferedWritesAtDeleteStart = -1;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    const createdTracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async () => {
+          deleteRunnerCalled = true;
+          bufferedWritesAtDeleteStart = (tracker as unknown as { queue: unknown[] }).queue.length;
+          await new Promise<void>((resolve) => {
+            deleteGate.release = resolve;
+          });
+        },
+      },
+    );
+    tracker = createdTracker;
+    createdTracker.handleMediaChange('/tmp/delete-yield-first.mkv', 'Delete Yield First');
+    createdTracker.handleMediaChange('/tmp/delete-yield-active.mkv', 'Delete Yield Active');
+
+    const privateApi = createdTracker as unknown as {
+      db: DatabaseSync;
+      queue: unknown[];
+      flushNow: () => void;
+    };
+    const sessionId = (
+      privateApi.db
+        .prepare(
+          `SELECT session_id AS sessionId
+           FROM imm_sessions
+           WHERE ended_at_ms IS NOT NULL
+           ORDER BY session_id
+           LIMIT 1`,
+        )
+        .get() as { sessionId: number } | null
+    )?.sessionId;
+    assert.ok(sessionId);
+
+    const deletePromise = createdTracker.deleteSession(sessionId);
+    let timerAdvanced = false;
+    setTimeout(() => {
+      timerAdvanced = true;
+    }, 0);
+
+    await waitForCondition(() => deleteRunnerCalled);
+    assert.equal(deleteRunnerCalled, true, 'delete should be dispatched to the maintenance runner');
+    assert.equal(
+      bufferedWritesAtDeleteStart,
+      0,
+      'writes buffered before delete should flush first',
+    );
+    await waitForCondition(() => timerAdvanced);
+
+    createdTracker.recordSubtitleLine('queued during delete', 0, 1);
+    privateApi.flushNow();
+    assert.ok(privateApi.queue.length > 0, 'tracking writes should wait for delete maintenance');
+
+    assert.ok(deleteGate.release);
+    deleteGate.release();
+    await deletePromise;
+  } finally {
+    deleteGate.release?.();
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('delete maintenance flushes the entire write queue before locking writes', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  const deleteGate: { release?: () => void } = {};
+  let queuedWritesAtDeleteStart = -1;
+  let writeLockedAtDeleteStart = false;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async () => {
+          const privateApi = tracker as unknown as {
+            queue: unknown[];
+            writeLock: { locked: boolean };
+          };
+          queuedWritesAtDeleteStart = privateApi.queue.length;
+          writeLockedAtDeleteStart = privateApi.writeLock.locked;
+          await new Promise<void>((resolve) => {
+            deleteGate.release = resolve;
+          });
+        },
+      },
+    );
+
+    const privateApi = tracker as unknown as {
+      batchSize: number;
+      flushNow: () => void;
+      queue: unknown[];
+    };
+    privateApi.batchSize = 1;
+    privateApi.queue.push({}, {}, {});
+    privateApi.flushNow = () => {
+      privateApi.queue.shift();
+    };
+
+    const deletePromise = tracker.deleteSession(101);
+    await waitForCondition(() => deleteGate.release !== undefined);
+
+    assert.equal(queuedWritesAtDeleteStart, 0);
+    assert.equal(writeLockedAtDeleteStart, true);
+
+    deleteGate.release?.();
+    await deletePromise;
+  } finally {
+    deleteGate.release?.();
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('delete maintenance tasks stay serialized under concurrent requests', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  const releases: Array<() => void> = [];
+  let activeTasks = 0;
+  let maxActiveTasks = 0;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async () => {
+          activeTasks += 1;
+          maxActiveTasks = Math.max(maxActiveTasks, activeTasks);
+          await new Promise<void>((resolve) => {
+            releases.push(resolve);
+          });
+          activeTasks -= 1;
+        },
+      },
+    );
+
+    const firstDelete = tracker.deleteSession(101);
+    await waitForCondition(() => releases.length === 1);
+    assert.equal(maxActiveTasks, 1);
+
+    const secondDelete = tracker.deleteSession(102);
+
+    releases[0]?.();
+    await waitForCondition(() => releases.length === 2);
+    assert.equal(maxActiveTasks, 1);
+
+    releases[1]?.();
+    await Promise.all([firstDelete, secondDelete]);
+  } finally {
+    for (const release of releases) release();
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('concurrent delete requests share one maintenance worker batch', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  const tasks: unknown[] = [];
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async (_path, task) => {
+          tasks.push(task);
+        },
+      },
+    );
+
+    const firstDelete = tracker.deleteSession(201);
+    const secondDelete = tracker.deleteSessions([202, 203]);
+    const thirdDelete = tracker.deleteVideo(204);
+    await Promise.all([firstDelete, secondDelete, thirdDelete]);
+
+    assert.equal(tasks.length, 1, 'concurrent deletes should use one maintenance pass');
+    assert.deepEqual(tasks[0], {
+      kind: 'batch',
+      tasks: [
+        { kind: 'session', sessionId: 201 },
+        { kind: 'sessions', sessionIds: [202, 203] },
+        { kind: 'video', videoId: 204 },
+      ],
+    });
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('destroy rejects delete requests waiting behind active maintenance', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  let releaseFirstTask: () => void = () => {};
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    let markFirstTaskStarted: () => void = () => {};
+    const firstTaskStarted = new Promise<void>((resolve) => {
+      markFirstTaskStarted = resolve;
+    });
+    tracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async () => {
+          markFirstTaskStarted();
+          await new Promise<void>((resolve) => {
+            releaseFirstTask = resolve;
+          });
+        },
+      },
+    );
+
+    const firstDelete = tracker.deleteSession(301);
+    await firstTaskStarted;
+    const queuedDelete = tracker.deleteSession(302);
+    tracker.destroy();
+
+    const queuedOutcome = await Promise.race([
+      queuedDelete.then(
+        () => 'resolved',
+        (error: unknown) =>
+          error instanceof Error && /shutting down/.test(error.message)
+            ? 'rejected'
+            : 'wrong-error',
+      ),
+      new Promise<'pending'>((resolve) => setTimeout(() => resolve('pending'), 25)),
+    ]);
+    assert.equal(queuedOutcome, 'rejected');
+    releaseFirstTask();
+    await firstDelete;
+  } finally {
+    releaseFirstTask();
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('delete requested after destroy rejects without running maintenance', async () => {
+  const dbPath = makeDbPath();
+  let maintenanceCalls = 0;
+  const Ctor = await loadTrackerCtor();
+  const tracker = new Ctor(
+    { dbPath },
+    {
+      runDeleteMaintenanceTask: async () => {
+        maintenanceCalls += 1;
+      },
+    },
+  );
+
+  tracker.destroy();
+
+  await assert.rejects(tracker.deleteSession(303), /shutting down/);
+  assert.equal(maintenanceCalls, 0);
+  cleanupDbPath(dbPath);
+});
+
+test('deleteSessions skips maintenance when no sessions are deletable', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  const tasks: unknown[] = [];
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async (_path, task) => {
+          tasks.push(task);
+        },
+      },
+    );
+
+    await tracker.deleteSessions([]);
+
+    assert.deepEqual(tasks, []);
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('queued video delete is skipped when that video becomes active before dispatch', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+  const tasks: Array<{ kind: string }> = [];
+  let releaseFirstTask: () => void = () => {};
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    const createdTracker = new Ctor(
+      { dbPath },
+      {
+        runDeleteMaintenanceTask: async (_path, task) => {
+          tasks.push(task);
+          if (tasks.length === 1) {
+            await new Promise<void>((resolve) => {
+              releaseFirstTask = resolve;
+            });
+          }
+        },
+      },
+    );
+    tracker = createdTracker;
+    createdTracker.handleMediaChange('/tmp/delete-race-target.mkv', 'Delete Race Target');
+    createdTracker.handleMediaChange('/tmp/delete-race-other.mkv', 'Delete Race Other');
+
+    const privateApi = createdTracker as unknown as { db: DatabaseSync };
+    const targetVideoId = (
+      privateApi.db
+        .prepare(`SELECT video_id AS videoId FROM imm_videos WHERE video_key LIKE '%target.mkv'`)
+        .get() as { videoId: number } | null
+    )?.videoId;
+    assert.ok(targetVideoId);
+
+    const firstDelete = createdTracker.deleteSession(999_001);
+    await waitForCondition(() => tasks.length === 1);
+
+    const queuedVideoDelete = createdTracker.deleteVideo(targetVideoId);
+    createdTracker.handleMediaChange('/tmp/delete-race-target.mkv', 'Delete Race Target');
+    releaseFirstTask();
+    await Promise.all([firstDelete, queuedVideoDelete]);
+
+    assert.deepEqual(
+      tasks.map((task) => task.kind),
+      ['session'],
+    );
+  } finally {
+    releaseFirstTask();
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
 test('deleteVideo ignores the currently active video and keeps new writes flushable', async () => {
   const dbPath = makeDbPath();
   let tracker: ImmersionTrackerService | null = null;
@@ -1785,6 +2132,78 @@ test('handleMediaChange reuses the same provisional anime row across matching fi
   }
 });
 
+test('local parsing reuses a unique compatible manual assignment from the same directory', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath });
+    const anchorPath = '/tmp/grouped/Incorrect Name S01E01.mkv';
+    tracker.handleMediaChange(anchorPath, 'Episode 1');
+    await waitForPendingAnimeMetadata(tracker);
+
+    const privateApi = tracker as unknown as {
+      db: DatabaseSync;
+      sessionState: { videoId: number } | null;
+    };
+    const anchorVideoId = privateApi.sessionState?.videoId;
+    assert.ok(anchorVideoId);
+    tracker.handleMediaChange(null, null);
+    const timestamp = toDbTimestamp(trackerNowMs());
+    const target = privateApi.db
+      .prepare(
+        `
+          INSERT INTO imm_anime (
+            normalized_title_key,
+            canonical_title,
+            CREATED_DATE,
+            LAST_UPDATE_DATE
+          ) VALUES ('correct show season 1', 'Correct Show Season 1', ?, ?)
+          RETURNING anime_id AS animeId
+        `,
+      )
+      .get(timestamp, timestamp) as { animeId: number };
+    await tracker.moveVideoToAnime(anchorVideoId, target.animeId);
+
+    tracker.handleMediaChange(anchorPath, 'Episode 1');
+    await waitForPendingAnimeMetadata(tracker);
+    tracker.handleMediaChange('/tmp/grouped/Another Wrong Name S01E02.mkv', 'Episode 2');
+    await waitForPendingAnimeMetadata(tracker);
+    tracker.handleMediaChange('/tmp/grouped/Another Wrong Name S02E01.mkv', 'Episode 1');
+    await waitForPendingAnimeMetadata(tracker);
+
+    const rows = privateApi.db
+      .prepare(
+        `
+          SELECT source_path AS sourcePath, anime_id AS animeId, anime_assignment_locked AS locked
+          FROM imm_videos
+          WHERE source_path LIKE '/tmp/grouped/%'
+          ORDER BY source_path
+        `,
+      )
+      .all() as Array<{ sourcePath: string; animeId: number; locked: number }>;
+    const assignments = new Map(rows.map((row) => [row.sourcePath, row]));
+    assert.deepEqual(assignments.get(anchorPath), {
+      sourcePath: anchorPath,
+      animeId: target.animeId,
+      locked: 1,
+    });
+    assert.deepEqual(assignments.get('/tmp/grouped/Another Wrong Name S01E02.mkv'), {
+      sourcePath: '/tmp/grouped/Another Wrong Name S01E02.mkv',
+      animeId: target.animeId,
+      locked: 0,
+    });
+    assert.notEqual(
+      assignments.get('/tmp/grouped/Another Wrong Name S02E01.mkv')?.animeId,
+      target.animeId,
+    );
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
 test('handleMediaChange splits matching parsed titles across distinct seasons', async () => {
   const dbPath = makeDbPath();
   let tracker: ImmersionTrackerService | null = null;
@@ -2271,6 +2690,67 @@ test('Jellyfin playback metadata links stream videos to existing series title',
   }
 });
 
+test('Jellyfin metadata refresh preserves a manual episode assignment', async () => {
+  const dbPath = makeDbPath();
+  let tracker: ImmersionTrackerService | null = null;
+
+  try {
+    const Ctor = await loadTrackerCtor();
+    tracker = new Ctor({ dbPath });
+    const metadata = {
+      mediaPath: 'http://jellyfin.local/Videos/item-locked/stream?api_key=token',
+      displayTitle: 'Parsed Show S01E01',
+      itemTitle: 'Episode 1',
+      seriesTitle: 'Parsed Show',
+      seasonNumber: 1,
+      episodeNumber: 1,
+      itemId: 'item-locked',
+    };
+    tracker.recordJellyfinPlaybackMetadata(metadata);
+
+    const privateApi = tracker as unknown as { db: DatabaseSync };
+    const video = privateApi.db.prepare('SELECT video_id AS videoId FROM imm_videos').get() as {
+      videoId: number;
+    };
+    const timestamp = toDbTimestamp(trackerNowMs());
+    const target = privateApi.db
+      .prepare(
+        `
+          INSERT INTO imm_anime (
+            normalized_title_key,
+            canonical_title,
+            CREATED_DATE,
+            LAST_UPDATE_DATE
+          ) VALUES ('correct show', 'Correct Show', ?, ?)
+          RETURNING anime_id AS animeId
+        `,
+      )
+      .get(timestamp, timestamp) as { animeId: number };
+
+    await tracker.moveVideoToAnime(video.videoId, target.animeId);
+    tracker.recordJellyfinPlaybackMetadata(metadata);
+
+    const assignment = privateApi.db
+      .prepare(
+        `
+          SELECT anime_id AS animeId, anime_assignment_locked AS locked
+          FROM imm_videos
+          WHERE video_id = ?
+        `,
+      )
+      .get(video.videoId) as { animeId: number; locked: number };
+    assert.equal(assignment.animeId, target.animeId);
+    assert.equal(assignment.locked, 1);
+    const animeCount = privateApi.db.prepare('SELECT COUNT(*) AS count FROM imm_anime').get() as {
+      count: number;
+    };
+    assert.equal(animeCount.count, 1);
+  } finally {
+    tracker?.destroy();
+    cleanupDbPath(dbPath);
+  }
+});
+
 test('startup repairs existing Jellyfin stream video links to metadata rows', async () => {
   const dbPath = makeDbPath();
   let tracker: ImmersionTrackerService | null = null;
@@ -3457,6 +3937,22 @@ test('reassignAnimeAnilist redistributes conflicting legacy combined row before
         (1, 2000, 1000, 1000, 1, 10, 0, 0, 0, 0, 0, 0, 0, 0),
         (2, 4000, 2000, 2000, 2, 20, 0, 0, 0, 0, 0, 0, 0, 0),
         (3, 6000, 3000, 3000, 3, 30, 0, 0, 0, 0, 0, 0, 0, 0);
+
+      -- The per-video lifetime rows those finalized sessions would have left
+      -- behind; redistributing videos re-derives imm_lifetime_anime from these.
+      INSERT INTO imm_lifetime_media (
+        video_id,
+        total_sessions,
+        total_active_ms,
+        completed,
+        first_watched_ms,
+        last_watched_ms,
+        CREATED_DATE,
+        LAST_UPDATE_DATE
+      ) VALUES
+        (1, 1, 1000, 0, '1000', '2000', 1000, 2000),
+        (2, 1, 2000, 0, '3000', '4000', 3000, 4000),
+        (3, 1, 3000, 0, '5000', '6000', 5000, 6000);
     `);
 
     await tracker.reassignAnimeAnilist(2, {

src/core/services/immersion-tracker-service.ts

@@ -16,6 +16,8 @@ import {
   applyPragmas,
   createTrackerPreparedStatements,
   ensureSchema,
+  findManualDirectoryAnimeAssignment,
+  getManualAnimeAssignment,
   executeQueuedWrite,
   getOrCreateAnimeRecord,
   getOrCreateVideoRecord,
@@ -28,6 +30,7 @@ import {
 } from './immersion-tracker/storage';
 import {
   applySessionLifetimeSummary,
+  recomputeLifetimeAnimeAggregates,
   reconcileStaleActiveSessions,
   rebuildLifetimeSummaries as rebuildLifetimeSummaryTables,
   shouldBackfillLifetimeSummaries,
@@ -83,19 +86,29 @@ import {
 } from './immersion-tracker/query-library';
 import {
   cleanupVocabularyStats,
-  deleteAnime as deleteAnimeQuery,
-  deleteSession as deleteSessionQuery,
-  deleteSessions as deleteSessionsQuery,
-  deleteVideo as deleteVideoQuery,
   getVideoDurationMs,
   markVideoWatched,
   upsertCoverArt,
 } from './immersion-tracker/query-maintenance';
+import {
+  DeleteMaintenanceWorkerRuntime,
+  type RunDeleteMaintenanceTask,
+} from './immersion-tracker/delete-maintenance-worker-runtime';
+import { DeleteMaintenanceScheduler } from './immersion-tracker/delete-maintenance-scheduler';
 import { repairJellyfinStreamVideoLinks } from './immersion-tracker/jellyfin-link-repair';
 import {
+  dismissAnimeMergeRecommendation,
+  getAnimeMergeRecommendations,
   repairLegacySeasonlessAnimeRows,
   resolveAnimeAnilistConflict,
+  type AnimeMergeRecommendation,
 } from './immersion-tracker/anime-season-repair';
+import {
+  mergeAnimeRecords,
+  moveVideoToAnime as moveVideoToAnimeQuery,
+  type AnimeMergeSummary,
+  type VideoMoveSummary,
+} from './immersion-tracker/anime-merge';
 import {
   buildVideoKey,
   deriveCanonicalTitle,
@@ -182,6 +195,7 @@ const YOUTUBE_SCREENSHOT_MAX_SECONDS = 120;
 const YOUTUBE_OEMBED_ENDPOINT = 'https://www.youtube.com/oembed';
 const YOUTUBE_ID_PATTERN = /^[A-Za-z0-9_-]{6,}$/;
 const YOUTUBE_METADATA_REFRESH_MS = 24 * 60 * 60 * 1000;
+const DELETE_MAINTENANCE_BATCH_WINDOW_MS = 10;
 
 function isValidYouTubeVideoId(value: string | null): boolean {
   return Boolean(value && YOUTUBE_ID_PATTERN.test(value));
@@ -385,6 +399,8 @@ export class ImmersionTrackerService {
   private readonly vacuumIntervalMs: number;
   private readonly dbPath: string;
   private readonly writeLock = { locked: false };
+  private readonly destroyDeleteMaintenanceRunner: () => void;
+  private readonly deleteMaintenanceScheduler: DeleteMaintenanceScheduler;
   private flushTimer: ReturnType<typeof setTimeout> | null = null;
   private maintenanceTimer: ReturnType<typeof setInterval> | null = null;
   private flushScheduled = false;
@@ -406,9 +422,37 @@ export class ImmersionTrackerService {
     | ((row: LegacyVocabularyPosRow) => Promise<LegacyVocabularyPosResolution | null>)
     | undefined;
 
-  constructor(options: ImmersionTrackerOptions) {
+  constructor(
+    options: ImmersionTrackerOptions,
+    dependencies: {
+      runDeleteMaintenanceTask?: RunDeleteMaintenanceTask;
+      destroyDeleteMaintenanceRunner?: () => void;
+    } = {},
+  ) {
     this.dbPath = options.dbPath;
     this.resolveLegacyVocabularyPos = options.resolveLegacyVocabularyPos;
+    let runDeleteMaintenanceTask: RunDeleteMaintenanceTask;
+    if (dependencies.runDeleteMaintenanceTask) {
+      runDeleteMaintenanceTask = dependencies.runDeleteMaintenanceTask;
+      this.destroyDeleteMaintenanceRunner =
+        dependencies.destroyDeleteMaintenanceRunner ?? (() => {});
+    } else {
+      const deleteMaintenanceRuntime = new DeleteMaintenanceWorkerRuntime();
+      runDeleteMaintenanceTask = (dbPath, task) => deleteMaintenanceRuntime.run(dbPath, task);
+      this.destroyDeleteMaintenanceRunner = () => deleteMaintenanceRuntime.destroy();
+    }
+    this.deleteMaintenanceScheduler = new DeleteMaintenanceScheduler({
+      batchWindowMs: DELETE_MAINTENANCE_BATCH_WINDOW_MS,
+      runTask: (task) => runDeleteMaintenanceTask(this.dbPath, task),
+      onBusy: () => {
+        this.requireWriteQueueDrained('delete maintenance');
+        this.writeLock.locked = true;
+      },
+      onIdle: () => {
+        this.writeLock.locked = false;
+        if (!this.isDestroyed && this.queue.length > 0) this.scheduleFlush(0);
+      },
+    });
     const parentDir = path.dirname(this.dbPath);
     if (!fs.existsSync(parentDir)) {
       fs.mkdirSync(parentDir, { recursive: true });
@@ -485,7 +529,7 @@ export class ImmersionTrackerService {
       this.logger.info(
         `Repaired season-scoped stats links on startup: scanned=${seasonRepair.scanned} movedVideos=${seasonRepair.movedVideos} deletedAnimeRows=${seasonRepair.deletedAnimeRows}`,
       );
-      rebuildLifetimeSummaryTables(this.db);
+      recomputeLifetimeAnimeAggregates(this.db);
     }
     if (shouldBackfillLifetimeSummaries(this.db)) {
       const result = rebuildLifetimeSummaryTables(this.db);
@@ -512,6 +556,8 @@ export class ImmersionTrackerService {
     }
     this.finalizeActiveSession();
     this.isDestroyed = true;
+    this.deleteMaintenanceScheduler.destroy();
+    this.destroyDeleteMaintenanceRunner();
     this.db.close();
   }
 
@@ -596,8 +642,7 @@ export class ImmersionTrackerService {
   }
 
   async rebuildLifetimeSummaries(): Promise<LifetimeRebuildSummary> {
-    this.flushTelemetry(true);
+    this.requireWriteQueueDrained('rebuilding lifetime summaries');
-    this.flushNow();
     return rebuildLifetimeSummaryTables(this.db);
   }
 
@@ -664,6 +709,14 @@ export class ImmersionTrackerService {
     return getAnimeLibrary(this.db);
   }
 
+  async getAnimeMergeRecommendations(): Promise<AnimeMergeRecommendation[]> {
+    return getAnimeMergeRecommendations(this.db);
+  }
+
+  async dismissAnimeMergeRecommendation(recommendationId: number): Promise<boolean> {
+    return dismissAnimeMergeRecommendation(this.db, recommendationId);
+  }
+
   async getAnimeDetail(animeId: number): Promise<AnimeDetailRow | null> {
     this.relinkYoutubeAnimeLibrary();
     return getAnimeDetail(this.db, animeId);
@@ -709,51 +762,123 @@ export class ImmersionTrackerService {
       this.logger.warn(`Ignoring delete request for active immersion session ${sessionId}`);
       return;
     }
-    deleteSessionQuery(this.db, sessionId);
+    await this.enqueueDeleteMaintenanceTask(() => ({ kind: 'session', sessionId }));
   }
 
   async deleteSessions(sessionIds: number[]): Promise<void> {
-    const activeSessionId = this.sessionState?.sessionId;
+    await this.enqueueDeleteMaintenanceTask(() => {
-    const deletableSessionIds =
+      const activeSessionId = this.sessionState?.sessionId;
-      activeSessionId === undefined
+      const deletableSessionIds =
-        ? sessionIds
+        activeSessionId === undefined
-        : sessionIds.filter((sessionId) => sessionId !== activeSessionId);
+          ? sessionIds
-    if (deletableSessionIds.length !== sessionIds.length) {
+          : sessionIds.filter((sessionId) => sessionId !== activeSessionId);
-      this.logger.warn(
+      if (deletableSessionIds.length !== sessionIds.length) {
-        `Ignoring bulk delete request for active immersion session ${activeSessionId}`,
+        this.logger.warn(
-      );
+          `Ignoring bulk delete request for active immersion session ${activeSessionId}`,
-    }
+        );
-    deleteSessionsQuery(this.db, deletableSessionIds);
+      }
+      if (deletableSessionIds.length === 0) return null;
+      return { kind: 'sessions', sessionIds: deletableSessionIds };
+    });
   }
 
   async deleteVideo(videoId: number): Promise<void> {
-    if (this.sessionState?.videoId === videoId) {
+    await this.enqueueDeleteMaintenanceTask(() => {
-      this.logger.warn(`Ignoring delete request for active immersion video ${videoId}`);
+      if (this.sessionState?.videoId === videoId) {
-      return;
+        this.logger.warn(`Ignoring delete request for active immersion video ${videoId}`);
-    }
+        return null;
-    deleteVideoQuery(this.db, videoId);
+      }
+      return { kind: 'video', videoId };
+    });
   }
 
   async deleteAnime(animeId: number): Promise<void> {
-    // The active video's anime link is assigned asynchronously after the title
+    await this.enqueueDeleteMaintenanceTask(async () => {
-    // is parsed, so a guard reading imm_videos too early sees a null and lets
+      // Resolve this at dispatch time because another queued delete can leave
-    // the delete through — then the late update recreates the anime row.
+      // enough time for playback to switch to an episode of this anime.
+      const pendingVideoId = this.sessionState?.videoId;
+      if (pendingVideoId !== undefined) {
+        await this.pendingAnimeMetadataUpdates.get(pendingVideoId);
+      }
+
+      const activeVideoId = this.sessionState?.videoId;
+      if (activeVideoId !== undefined) {
+        const activeAnime = this.db
+          .prepare('SELECT anime_id FROM imm_videos WHERE video_id = ?')
+          .get(activeVideoId) as { anime_id: number | null } | null;
+        if (activeAnime?.anime_id === animeId) {
+          this.logger.warn(`Ignoring delete request for active immersion anime ${animeId}`);
+          return null;
+        }
+      }
+      return { kind: 'anime', animeId };
+    });
+  }
+
+  private enqueueDeleteMaintenanceTask(
+    resolveTask: Parameters<DeleteMaintenanceScheduler['enqueue']>[0],
+  ): Promise<void> {
+    if (this.isDestroyed) {
+      return Promise.reject(new Error('Immersion tracker is shutting down'));
+    }
+    return this.deleteMaintenanceScheduler.enqueue(resolveTask);
+  }
+
+  /**
+   * Fold duplicate library entries into one. Sources that hold the currently
+   * playing episode are fine: the videos move, nothing is deleted out from
+   * under the active session.
+   */
+  async mergeAnime(targetAnimeId: number, sourceAnimeIds: number[]): Promise<AnimeMergeSummary> {
     const pendingVideoId = this.sessionState?.videoId;
     if (pendingVideoId !== undefined) {
       await this.pendingAnimeMetadataUpdates.get(pendingVideoId);
     }
+    // This rebuilds the lifetime summaries, which recompute from the database:
+    // queued writes have to land first or the active session is dropped from
+    // the merged totals.
+    this.requireWriteQueueDrained('merging library entries');
+    return mergeAnimeRecords(this.db, targetAnimeId, sourceAnimeIds);
+  }
 
-    const activeVideoId = this.sessionState?.videoId;
+  async moveVideoToAnime(videoId: number, targetAnimeId: number): Promise<VideoMoveSummary> {
-    if (activeVideoId !== undefined) {
+    await this.pendingAnimeMetadataUpdates.get(videoId);
-      const activeAnime = this.db
+    this.requireWriteQueueDrained('moving an episode');
-        .prepare('SELECT anime_id FROM imm_videos WHERE video_id = ?')
+    return moveVideoToAnimeQuery(this.db, videoId, targetAnimeId);
-        .get(activeVideoId) as { anime_id: number | null } | null;
+  }
-      if (activeAnime?.anime_id === animeId) {
+
-        this.logger.warn(`Ignoring delete request for active immersion anime ${animeId}`);
+  /**
-        return;
+   * Persist every queued write before a caller recomputes summaries from the
+   * database.
+   *
+   * A single `flushNow()` is not enough: forced telemetry is appended to the
+   * back of the queue while `flushNow()` writes at most `batchSize` entries off
+   * the front, so a busy session leaves the newest sample unwritten. Stops as
+   * soon as a pass makes no progress — a rolled-back batch is pushed back onto
+   * the queue, and looping on that would spin forever.
+   *
+   * Returns false when the queue could not be emptied. Summary-rebuilding
+   * callers fail closed in that case.
+   */
+  private drainWriteQueue(context: string): boolean {
+    this.flushTelemetry(true);
+    while (this.queue.length > 0) {
+      const pending = this.queue.length;
+      this.flushNow();
+      if (this.queue.length >= pending) {
+        this.logger.warn(
+          `Immersion tracker queue did not drain before ${context}; summaries may lag by ${this.queue.length} writes`,
+        );
+        return false;
       }
     }
-    deleteAnimeQuery(this.db, animeId);
+    return true;
+  }
+
+  private requireWriteQueueDrained(context: string): void {
+    if (!this.drainWriteQueue(context)) {
+      throw new Error(`Immersion tracker queue did not drain before ${context}`);
+    }
   }
 
   async reassignAnimeAnilist(
@@ -768,7 +893,14 @@ export class ImmersionTrackerService {
       coverUrl?: string | null;
     },
   ): Promise<void> {
-    const repair = resolveAnimeAnilistConflict(this.db, animeId, info.anilistId);
+    this.requireWriteQueueDrained('reassigning an AniList entry');
+    // The user is acting on this entry, so it is the one that survives when
+    // another row already claims the same AniList id.
+    const repair = resolveAnimeAnilistConflict(this.db, animeId, info.anilistId, {
+      survivor: 'target',
+      matchConfidence: 'manual',
+    });
+    if (repair.anilistAssignmentBlocked) return;
     this.db
       .prepare(
         `
@@ -795,7 +927,7 @@ export class ImmersionTrackerService {
         animeId,
       );
     if (repair.movedVideos > 0 || repair.deletedAnimeRows > 0) {
-      rebuildLifetimeSummaryTables(this.db);
+      recomputeLifetimeAnimeAggregates(this.db);
     }
 
     // Update cover art for all videos in this anime
@@ -1243,7 +1375,7 @@ export class ImmersionTrackerService {
         metadataJson: candidate.metadataJson,
       });
     }
-    rebuildLifetimeSummaryTables(this.db);
+    recomputeLifetimeAnimeAggregates(this.db);
   }
 
   recordJellyfinPlaybackMetadata(metadata: JellyfinPlaybackMetadataInput): void {
@@ -1291,16 +1423,18 @@ export class ImmersionTrackerService {
       seasonNumber,
       episodeNumber,
     });
-    const animeId = getOrCreateAnimeRecord(this.db, {
+    const animeId =
-      parsedTitle: libraryTitle,
+      getManualAnimeAssignment(this.db, videoId) ??
-      canonicalTitle: libraryTitle,
+      getOrCreateAnimeRecord(this.db, {
-      seasonScope: seasonNumber,
+        parsedTitle: libraryTitle,
-      anilistId: null,
+        canonicalTitle: libraryTitle,
-      titleRomaji: null,
+        seasonScope: seasonNumber,
-      titleEnglish: null,
+        anilistId: null,
-      titleNative: null,
+        titleRomaji: null,
-      metadataJson,
+        titleEnglish: null,
-    });
+        titleNative: null,
+        metadataJson,
+      });
     linkVideoToAnimeRecord(this.db, videoId, {
       animeId,
       parsedBasename: null,
@@ -1316,7 +1450,7 @@ export class ImmersionTrackerService {
       this.db.prepare('SELECT 1 FROM imm_lifetime_media WHERE video_id = ?').get(videoId),
     );
     if (hasLifetimeMedia || (previousLink && previousLink.animeId !== animeId)) {
-      rebuildLifetimeSummaryTables(this.db);
+      recomputeLifetimeAnimeAggregates(this.db);
     }
   }
 
@@ -1811,7 +1945,7 @@ export class ImmersionTrackerService {
   }
 
   private runMaintenance(): void {
-    if (this.isDestroyed) return;
+    if (this.isDestroyed || this.writeLock.locked) return;
     try {
       this.flushTelemetry(true);
       this.flushNow();
@@ -1937,16 +2071,21 @@ export class ImmersionTrackerService {
           return;
         }
 
-        const animeId = getOrCreateAnimeRecord(this.db, {
+        const animeId =
-          parsedTitle: parsed.parsedTitle,
+          getManualAnimeAssignment(this.db, videoId) ??
-          canonicalTitle: parsed.parsedTitle,
+          (mediaPath && !isRemoteSource(mediaPath)
-          seasonScope: parsed.parsedSeason,
+            ? findManualDirectoryAnimeAssignment(this.db, videoId, mediaPath, parsed.parsedSeason)
-          anilistId: null,
+            : null) ??
-          titleRomaji: null,
+          getOrCreateAnimeRecord(this.db, {
-          titleEnglish: null,
+            parsedTitle: parsed.parsedTitle,
-          titleNative: null,
+            canonicalTitle: parsed.parsedTitle,
-          metadataJson: parsed.parseMetadataJson,
+            seasonScope: parsed.parsedSeason,
-        });
+            anilistId: null,
+            titleRomaji: null,
+            titleEnglish: null,
+            titleNative: null,
+            metadataJson: parsed.parseMetadataJson,
+          });
         linkVideoToAnimeRecord(this.db, videoId, {
           animeId,
           parsedBasename: parsed.parsedBasename,

src/core/services/immersion-tracker/__tests__/anime-merge.test.ts

@@ -0,0 +1,896 @@
+import assert from 'node:assert/strict';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import test from 'node:test';
+import { Database } from '../sqlite.js';
+import type { DatabaseSync } from '../sqlite.js';
+import {
+  applyPragmas,
+  ensureSchema,
+  findManualDirectoryAnimeAssignment,
+  getManualAnimeAssignment,
+  getOrCreateAnimeRecord,
+  linkVideoToAnimeRecord,
+} from '../storage.js';
+import { mergeAnimeRecords, moveVideoToAnime } from '../anime-merge.js';
+import {
+  dismissAnimeMergeRecommendation,
+  getAnimeMergeRecommendations,
+  resolveAnimeAnilistConflict,
+} from '../anime-season-repair.js';
+import { updateAnimeAnilistInfo } from '../query-maintenance.js';
+
+const BASE_MS = 1_700_000_000_000;
+
+function makeDbPath(): string {
+  const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'subminer-anime-merge-test-'));
+  return path.join(dir, 'immersion.sqlite');
+}
+
+function cleanupDbPath(dbPath: string): void {
+  const dir = path.dirname(dbPath);
+  if (!fs.existsSync(dir)) return;
+  fs.rmSync(dir, { recursive: true, force: true });
+}
+
+function withDb(work: (db: DatabaseSync) => void): void {
+  const dbPath = makeDbPath();
+  const db = new Database(dbPath);
+  try {
+    applyPragmas(db);
+    ensureSchema(db);
+    work(db);
+  } finally {
+    db.close();
+    cleanupDbPath(dbPath);
+  }
+}
+
+interface AnimeSeed {
+  animeId: number;
+  key: string;
+  title: string;
+  anilistId?: number | null;
+  titleRomaji?: string | null;
+}
+
+function insertAnime(db: DatabaseSync, seed: AnimeSeed): void {
+  db.prepare(
+    `INSERT INTO imm_anime(anime_id, normalized_title_key, canonical_title, anilist_id, title_romaji, CREATED_DATE, LAST_UPDATE_DATE)
+     VALUES (?, ?, ?, ?, ?, ?, ?)`,
+  ).run(
+    seed.animeId,
+    seed.key,
+    seed.title,
+    seed.anilistId ?? null,
+    seed.titleRomaji ?? null,
+    BASE_MS,
+    BASE_MS,
+  );
+}
+
+interface EpisodeSeed {
+  videoId: number;
+  animeId: number;
+  season?: number | null;
+  episode?: number;
+  activeMs?: number;
+  cards?: number;
+}
+
+/**
+ * One episode with one ended session, plus the imm_lifetime_media row the
+ * session would have left behind, so lifetime aggregates have something to sum.
+ */
+function insertEpisode(db: DatabaseSync, seed: EpisodeSeed): void {
+  const activeMs = seed.activeMs ?? 1000;
+  const cards = seed.cards ?? 1;
+  db.prepare(
+    `INSERT INTO imm_videos(video_id, video_key, anime_id, canonical_title, source_type, parsed_title, parsed_season, parsed_episode, watched, duration_ms, CREATED_DATE, LAST_UPDATE_DATE)
+     VALUES (?, ?, ?, ?, 1, 'Show', ?, ?, 1, 1440000, ?, ?)`,
+  ).run(
+    seed.videoId,
+    `local:/tmp/show-${seed.videoId}.mkv`,
+    seed.animeId,
+    `Show ${seed.videoId}`,
+    seed.season ?? null,
+    seed.episode ?? seed.videoId,
+    BASE_MS,
+    BASE_MS,
+  );
+  db.prepare(
+    `INSERT INTO imm_sessions(session_id, session_uuid, video_id, started_at_ms, ended_at_ms, status, active_watched_ms, cards_mined, CREATED_DATE, LAST_UPDATE_DATE)
+     VALUES (?, ?, ?, ?, ?, 2, ?, ?, ?, ?)`,
+  ).run(
+    seed.videoId,
+    `session-${seed.videoId}`,
+    seed.videoId,
+    String(BASE_MS),
+    String(BASE_MS + activeMs),
+    activeMs,
+    cards,
+    BASE_MS,
+    BASE_MS,
+  );
+  db.prepare(
+    `INSERT INTO imm_subtitle_lines(session_id, video_id, anime_id, line_index, text, CREATED_DATE, LAST_UPDATE_DATE)
+     VALUES (?, ?, ?, 1, ?, ?, ?)`,
+  ).run(seed.videoId, seed.videoId, seed.animeId, `line ${seed.videoId}`, BASE_MS, BASE_MS);
+  db.prepare(
+    `INSERT INTO imm_lifetime_media(video_id, total_sessions, total_active_ms, total_cards, completed, first_watched_ms, last_watched_ms, CREATED_DATE, LAST_UPDATE_DATE)
+     VALUES (?, 1, ?, ?, 1, ?, ?, ?, ?)`,
+  ).run(
+    seed.videoId,
+    activeMs,
+    cards,
+    String(BASE_MS),
+    String(BASE_MS + activeMs),
+    BASE_MS,
+    BASE_MS,
+  );
+}
+
+function animeIds(db: DatabaseSync): number[] {
+  return (
+    db.prepare('SELECT anime_id AS id FROM imm_anime ORDER BY anime_id').all() as Array<{
+      id: number;
+    }>
+  ).map((row) => row.id);
+}
+
+function videoAnimeId(db: DatabaseSync, videoId: number): number | null {
+  return (
+    db.prepare('SELECT anime_id AS id FROM imm_videos WHERE video_id = ?').get(videoId) as {
+      id: number | null;
+    }
+  ).id;
+}
+
+function assignmentLocked(db: DatabaseSync, videoId: number): number {
+  return (
+    db
+      .prepare('SELECT anime_assignment_locked AS locked FROM imm_videos WHERE video_id = ?')
+      .get(videoId) as { locked: number }
+  ).locked;
+}
+
+function lineAnimeIds(db: DatabaseSync, animeId: number): number {
+  return Number(
+    (
+      db
+        .prepare('SELECT COUNT(*) AS total FROM imm_subtitle_lines WHERE anime_id = ?')
+        .get(animeId) as { total: number }
+    ).total,
+  );
+}
+
+test('mergeAnimeRecords folds episodes, lines and lifetime totals into the target', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1', anilistId: 555 });
+    insertEpisode(db, { videoId: 1, animeId: 1, activeMs: 1000, cards: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1, activeMs: 2000, cards: 3 });
+
+    const summary = mergeAnimeRecords(db, 1, [2]);
+
+    assert.equal(summary.survivingAnimeId, 1);
+    assert.deepEqual(summary.mergedAnimeIds, [2]);
+    assert.equal(summary.movedVideos, 1);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(videoAnimeId(db, 2), 1);
+    assert.equal(lineAnimeIds(db, 1), 2);
+
+    const lifetime = db
+      .prepare(
+        'SELECT total_active_ms AS activeMs, total_cards AS cards, episodes_started AS episodes FROM imm_lifetime_anime WHERE anime_id = 1',
+      )
+      .get() as { activeMs: number; cards: number; episodes: number };
+    assert.equal(lifetime.activeMs, 3000);
+    assert.equal(lifetime.cards, 4);
+    assert.equal(lifetime.episodes, 2);
+  });
+});
+
+test('merge and move preserve lifetime history whose raw sessions were pruned', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertAnime(db, { animeId: 3, key: 'other show', title: 'Other Show' });
+    insertEpisode(db, { videoId: 1, animeId: 1, activeMs: 1000, cards: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1, activeMs: 2000, cards: 3 });
+    insertEpisode(db, { videoId: 3, animeId: 3, activeMs: 4000, cards: 5 });
+    // Retention pruned every raw session; only the lifetime summaries remain.
+    db.exec('DELETE FROM imm_sessions');
+    db.prepare(
+      `UPDATE imm_lifetime_global
+       SET total_sessions = 200, total_active_ms = 360000000, total_cards = 500, active_days = 90
+       WHERE global_id = 1`,
+    ).run();
+
+    mergeAnimeRecords(db, 1, [2]);
+    moveVideoToAnime(db, 3, 1);
+
+    const globalRow = db
+      .prepare(
+        `SELECT total_sessions AS sessions, total_active_ms AS activeMs, total_cards AS cards, active_days AS days
+         FROM imm_lifetime_global WHERE global_id = 1`,
+      )
+      .get() as { sessions: number; activeMs: number; cards: number; days: number };
+    assert.equal(globalRow.sessions, 200);
+    assert.equal(globalRow.activeMs, 360000000);
+    assert.equal(globalRow.cards, 500);
+    assert.equal(globalRow.days, 90);
+
+    const survivor = db
+      .prepare(
+        `SELECT total_active_ms AS activeMs, total_cards AS cards, episodes_started AS episodes
+         FROM imm_lifetime_anime WHERE anime_id = 1`,
+      )
+      .get() as { activeMs: number; cards: number; episodes: number };
+    assert.equal(survivor.activeMs, 7000);
+    assert.equal(survivor.cards, 9);
+    assert.equal(survivor.episodes, 3);
+    assert.equal(
+      db.prepare('SELECT 1 FROM imm_lifetime_anime WHERE anime_id = 3').get(),
+      undefined,
+    );
+  });
+});
+
+test('mergeAnimeRecords repoints subtitle lines recorded before the anime link landed', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+    // Lines are written with the video's anime_id at the time, which is NULL
+    // until the async title parse assigns one.
+    db.prepare(
+      `INSERT INTO imm_subtitle_lines(session_id, video_id, anime_id, line_index, text, CREATED_DATE, LAST_UPDATE_DATE)
+       VALUES (2, 2, NULL, 2, 'unlinked line', ?, ?)`,
+    ).run(BASE_MS, BASE_MS);
+
+    mergeAnimeRecords(db, 1, [2]);
+
+    assert.equal(lineAnimeIds(db, 1), 3);
+    const orphaned = Number(
+      (
+        db
+          .prepare('SELECT COUNT(*) AS total FROM imm_subtitle_lines WHERE anime_id IS NULL')
+          .get() as { total: number }
+      ).total,
+    );
+    assert.equal(orphaned, 0);
+  });
+});
+
+test('mergeAnimeRecords inherits metadata the target is missing without clobbering its own', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', titleRomaji: 'Shou' });
+    insertAnime(db, {
+      animeId: 2,
+      key: 'show season 1',
+      title: 'Show Season 1',
+      anilistId: 555,
+      titleRomaji: 'Show Romaji',
+    });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    mergeAnimeRecords(db, 1, [2]);
+
+    const row = db
+      .prepare(
+        'SELECT canonical_title AS title, anilist_id AS anilistId, title_romaji AS romaji FROM imm_anime WHERE anime_id = 1',
+      )
+      .get() as { title: string; anilistId: number | null; romaji: string | null };
+    assert.equal(row.title, 'Show');
+    // anilist_id is UNIQUE, so inheriting it proves the source row was gone first.
+    assert.equal(row.anilistId, 555);
+    assert.equal(row.romaji, 'Shou');
+  });
+});
+
+test('mergeAnimeRecords preserves source title identities as aliases of the survivor', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    db.prepare(
+      `INSERT INTO imm_anime_title_aliases(normalized_title_key, anime_id, CREATED_DATE, LAST_UPDATE_DATE)
+       VALUES ('show s01', 2, ?, ?)`,
+    ).run(BASE_MS, BASE_MS);
+
+    mergeAnimeRecords(db, 1, [2]);
+
+    const fromSourceTitle = getOrCreateAnimeRecord(db, {
+      parsedTitle: 'Show Season 1',
+      canonicalTitle: 'Show Season 1',
+      seasonScope: 1,
+      anilistId: null,
+      titleRomaji: null,
+      titleEnglish: null,
+      titleNative: null,
+      metadataJson: null,
+    });
+    const fromTransferredAlias = getOrCreateAnimeRecord(db, {
+      parsedTitle: 'Show S01',
+      canonicalTitle: 'Show S01',
+      anilistId: null,
+      titleRomaji: null,
+      titleEnglish: null,
+      titleNative: null,
+      metadataJson: null,
+    });
+
+    assert.equal(fromSourceTitle, 1);
+    assert.equal(fromTransferredAlias, 1);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(
+      (
+        db.prepare('SELECT canonical_title AS title FROM imm_anime WHERE anime_id = 1').get() as {
+          title: string;
+        }
+      ).title,
+      'Show',
+    );
+  });
+});
+
+test('mergeAnimeRecords ignores unknown targets and self-merges', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+
+    assert.deepEqual(mergeAnimeRecords(db, 99, [1]).mergedAnimeIds, []);
+    assert.deepEqual(mergeAnimeRecords(db, 1, [1]).mergedAnimeIds, []);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(videoAnimeId(db, 1), 1);
+  });
+});
+
+test('moveVideoToAnime moves one episode and prunes the emptied entry', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'stray', title: 'Stray Episode Title', anilistId: 777 });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, activeMs: 5000, cards: 2 });
+
+    const summary = moveVideoToAnime(db, 2, 1);
+
+    assert.equal(summary.targetAnimeId, 1);
+    assert.equal(summary.previousAnimeId, 2);
+    assert.equal(summary.removedPreviousAnime, true);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(videoAnimeId(db, 2), 1);
+    assert.equal(assignmentLocked(db, 2), 1);
+    assert.equal(getManualAnimeAssignment(db, 2), 1);
+    assert.equal(lineAnimeIds(db, 1), 2);
+    const lifetime = db
+      .prepare('SELECT total_active_ms AS activeMs FROM imm_lifetime_anime WHERE anime_id = 1')
+      .get() as { activeMs: number };
+    assert.equal(lifetime.activeMs, 6000);
+    // The stray entry's AniList link is dropped, not inherited: a move makes no
+    // claim that the two entries are the same show.
+    const target = db
+      .prepare('SELECT anilist_id AS anilistId FROM imm_anime WHERE anime_id = 1')
+      .get() as { anilistId: number | null };
+    assert.equal(target.anilistId, null);
+  });
+});
+
+test('moveVideoToAnime is a no-op when the episode is already in the target entry', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+
+    const summary = moveVideoToAnime(db, 1, 1);
+
+    assert.equal(summary.targetAnimeId, 1);
+    assert.equal(summary.previousAnimeId, 1);
+    assert.equal(summary.removedPreviousAnime, false);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(videoAnimeId(db, 1), 1);
+    assert.equal(assignmentLocked(db, 1), 1);
+  });
+});
+
+test('automatic metadata cannot overwrite a manual episode assignment', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'stray', title: 'Stray' });
+    insertAnime(db, { animeId: 3, key: 'parser result', title: 'Parser Result' });
+    insertEpisode(db, { videoId: 1, animeId: 2, season: 1 });
+
+    moveVideoToAnime(db, 1, 1);
+    linkVideoToAnimeRecord(db, 1, {
+      animeId: 3,
+      parsedBasename: 'Parser Result S01E01.mkv',
+      parsedTitle: 'Parser Result',
+      parsedSeason: 1,
+      parsedEpisode: 1,
+      parserSource: 'guessit',
+      parserConfidence: 1,
+      parseMetadataJson: null,
+    });
+
+    assert.equal(videoAnimeId(db, 1), 1);
+    assert.equal(getManualAnimeAssignment(db, 1), 1);
+    const parsedTitle = db
+      .prepare('SELECT parsed_title AS parsedTitle FROM imm_videos WHERE video_id = 1')
+      .get() as { parsedTitle: string | null };
+    assert.equal(parsedTitle.parsedTitle, 'Parser Result');
+  });
+});
+
+test('directory grouping requires one season-compatible manual destination', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'stray', title: 'Stray' });
+    insertAnime(db, { animeId: 3, key: 'other', title: 'Other' });
+    insertEpisode(db, { videoId: 1, animeId: 2, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 3, season: 1 });
+    insertEpisode(db, { videoId: 3, animeId: 3, season: 1 });
+    db.prepare('UPDATE imm_videos SET source_path = ? WHERE video_id = ?').run(
+      '/library/show/Show S01E01.mkv',
+      1,
+    );
+    db.prepare('UPDATE imm_videos SET source_path = ? WHERE video_id = ?').run(
+      '/library/show/Stray S01E02.mkv',
+      2,
+    );
+    db.prepare('UPDATE imm_videos SET source_path = ? WHERE video_id = ?').run(
+      '/library/show/Other S01E03.mkv',
+      3,
+    );
+
+    moveVideoToAnime(db, 1, 1);
+
+    assert.equal(findManualDirectoryAnimeAssignment(db, 2, '/library/show/Stray S01E02.mkv', 1), 1);
+    assert.equal(
+      findManualDirectoryAnimeAssignment(db, 2, '/library/show/Stray S02E02.mkv', 2),
+      null,
+    );
+    assert.equal(
+      findManualDirectoryAnimeAssignment(db, 2, '/library/other/Stray S01E02.mkv', 1),
+      null,
+    );
+
+    moveVideoToAnime(db, 3, 3);
+    assert.equal(
+      findManualDirectoryAnimeAssignment(db, 2, '/library/show/Stray S01E02.mkv', 1),
+      null,
+    );
+  });
+});
+
+test('moveVideoToAnime keeps the source entry when other episodes remain', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertAnime(db, { animeId: 2, key: 'other', title: 'Other' });
+    insertEpisode(db, { videoId: 1, animeId: 2 });
+    insertEpisode(db, { videoId: 2, animeId: 2 });
+
+    const summary = moveVideoToAnime(db, 2, 1);
+
+    assert.equal(summary.removedPreviousAnime, false);
+    assert.deepEqual(animeIds(db), [1, 2]);
+    assert.equal(videoAnimeId(db, 1), 2);
+    assert.equal(videoAnimeId(db, 2), 1);
+  });
+});
+
+test('moveVideoToAnime rejects unknown episodes and targets', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+
+    assert.throws(() => moveVideoToAnime(db, 99, 1));
+    assert.throws(() => moveVideoToAnime(db, 1, 99));
+    assert.equal(videoAnimeId(db, 1), 1);
+  });
+});
+
+test('resolveAnimeAnilistConflict folds a seasonless duplicate into the entry that owns the id', () => {
+  withDb((db) => {
+    // Same show, split because one release tagged S01 and the other did not.
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', anilistId: 163132 });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132);
+
+    assert.equal(summary.survivingAnimeId, 1);
+    assert.equal(summary.movedVideos, 1);
+    assert.equal(summary.deletedAnimeRows, 1);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(videoAnimeId(db, 2), 1);
+  });
+});
+
+test('resolveAnimeAnilistConflict recommends a weak title collision instead of merging it', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'actual show',
+      title: 'Actual Show',
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'unrelated release', title: 'Unrelated Release' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132);
+
+    assert.equal(summary.repaired, 0);
+    assert.deepEqual(animeIds(db), [1, 2]);
+    assert.equal(videoAnimeId(db, 2), 2);
+    assert.deepEqual(getAnimeMergeRecommendations(db), [{ recommendationId: 1, animeIds: [1, 2] }]);
+  });
+});
+
+test('automatic AniList update leaves a weak collision unassigned for user review', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'actual show',
+      title: 'Actual Show',
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'unrelated release', title: 'Unrelated Release' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    updateAnimeAnilistInfo(db, 2, {
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+      titleEnglish: null,
+      titleNative: null,
+      episodesTotal: 12,
+      exactTitleMatch: false,
+    });
+
+    const target = db
+      .prepare('SELECT anilist_id AS anilistId FROM imm_anime WHERE anime_id = 2')
+      .get() as {
+      anilistId: number | null;
+    };
+    assert.equal(target.anilistId, null);
+    assert.deepEqual(getAnimeMergeRecommendations(db), [{ recommendationId: 1, animeIds: [1, 2] }]);
+  });
+});
+
+test('dismissed weak collision stays dismissed when automatic resolution repeats', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'actual show',
+      title: 'Actual Show',
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'unrelated release', title: 'Unrelated Release' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    resolveAnimeAnilistConflict(db, 2, 163132);
+    assert.equal(dismissAnimeMergeRecommendation(db, 1), true);
+    resolveAnimeAnilistConflict(db, 2, 163132);
+
+    assert.deepEqual(getAnimeMergeRecommendations(db), []);
+  });
+});
+
+test('dismissed recommendation prevents a later exact automatic merge of the pair', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'actual show',
+      title: 'Actual Show',
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'unrelated release', title: 'Unrelated Release' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    resolveAnimeAnilistConflict(db, 2, 163132, { matchConfidence: 'weak' });
+    assert.equal(dismissAnimeMergeRecommendation(db, 1), true);
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132, { matchConfidence: 'exact' });
+
+    assert.equal(summary.repaired, 0);
+    assert.deepEqual(animeIds(db), [1, 2]);
+    assert.equal(videoAnimeId(db, 2), 2);
+    assert.deepEqual(getAnimeMergeRecommendations(db), []);
+  });
+});
+
+test('manual merge clears recommendations involving the absorbed entry', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'actual show',
+      title: 'Actual Show',
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'unrelated release', title: 'Unrelated Release' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    resolveAnimeAnilistConflict(db, 2, 163132);
+    mergeAnimeRecords(db, 1, [2]);
+
+    assert.deepEqual(getAnimeMergeRecommendations(db), []);
+  });
+});
+
+test('resolveAnimeAnilistConflict keeps the target entry when the user drove the change', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', anilistId: 163132 });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132, { survivor: 'target' });
+
+    assert.equal(summary.survivingAnimeId, 2);
+    assert.deepEqual(animeIds(db), [2]);
+    assert.equal(videoAnimeId(db, 1), 2);
+    const row = db.prepare('SELECT anilist_id AS id FROM imm_anime WHERE anime_id = 2').get() as {
+      id: number | null;
+    };
+    assert.equal(row.id, 163132);
+  });
+});
+
+test('resolveAnimeAnilistConflict falls back to season redistribution for multi-season rows', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', anilistId: 163132 });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 1, season: 2 });
+    insertEpisode(db, { videoId: 3, animeId: 2, season: 1 });
+
+    resolveAnimeAnilistConflict(db, 2, 163132);
+
+    // The mixed row is split by season instead of being poured onto one card.
+    const titles = (
+      db.prepare('SELECT canonical_title AS title FROM imm_anime ORDER BY title').all() as Array<{
+        title: string;
+      }>
+    ).map((row) => row.title);
+    assert.deepEqual(titles, ['Show Season 1', 'Show Season 2']);
+    assert.equal(videoAnimeId(db, 1), 2);
+    assert.equal(videoAnimeId(db, 3), 2);
+    assert.notEqual(videoAnimeId(db, 2), 2);
+  });
+});
+
+test('season redistribution leaves manually assigned episodes in place', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', anilistId: 163132 });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 1, season: 2 });
+    insertEpisode(db, { videoId: 3, animeId: 2, season: 1 });
+    moveVideoToAnime(db, 1, 1);
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132);
+
+    assert.equal(videoAnimeId(db, 1), 1);
+    assert.equal(assignmentLocked(db, 1), 1);
+    assert.notEqual(videoAnimeId(db, 2), 1);
+    assert.equal(summary.movedVideos, 1);
+  });
+});
+
+test('resolveAnimeAnilistConflict leaves explicit incompatible seasons and assignments unchanged', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'show season 1',
+      title: 'Show Season 1',
+      anilistId: 163132,
+      titleRomaji: 'Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'show season 2', title: 'Show Season 2' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 2 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132, { matchConfidence: 'exact' });
+
+    assert.equal(summary.repaired, 0);
+    assert.equal(summary.movedVideos, 0);
+    assert.equal(summary.deletedAnimeRows, 0);
+    assert.deepEqual(animeIds(db), [1, 2]);
+    assert.equal(videoAnimeId(db, 1), 1);
+    assert.equal(videoAnimeId(db, 2), 2);
+    const assignments = db
+      .prepare(
+        'SELECT anime_id AS animeId, anilist_id AS anilistId FROM imm_anime ORDER BY anime_id',
+      )
+      .all() as Array<{ animeId: number; anilistId: number | null }>;
+    assert.deepEqual(assignments, [
+      { animeId: 1, anilistId: 163132 },
+      { animeId: 2, anilistId: null },
+    ]);
+    assert.deepEqual(getAnimeMergeRecommendations(db), []);
+  });
+});
+
+test('manual AniList resolution reassigns across explicit seasons without merging them', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'show season 1',
+      title: 'Show Season 1',
+      anilistId: 163132,
+      titleRomaji: 'Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'show season 2', title: 'Show Season 2' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 2 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132, { survivor: 'target' });
+
+    assert.equal(summary.anilistAssignmentBlocked, false);
+    assert.deepEqual(animeIds(db), [1, 2]);
+    const assignments = db
+      .prepare(
+        'SELECT anime_id AS animeId, anilist_id AS anilistId FROM imm_anime ORDER BY anime_id',
+      )
+      .all() as Array<{ animeId: number; anilistId: number | null }>;
+    assert.deepEqual(assignments, [
+      { animeId: 1, anilistId: null },
+      { animeId: 2, anilistId: 163132 },
+    ]);
+  });
+});
+
+test('automatic AniList update does not transfer an assignment across explicit seasons', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'show season 1',
+      title: 'Show Season 1',
+      anilistId: 163132,
+      titleRomaji: 'Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'show season 2', title: 'Show Season 2' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 2 });
+
+    updateAnimeAnilistInfo(db, 2, {
+      anilistId: 163132,
+      titleRomaji: 'Show',
+      titleEnglish: null,
+      titleNative: null,
+      episodesTotal: 12,
+      exactTitleMatch: true,
+    });
+
+    const assignments = db
+      .prepare(
+        'SELECT anime_id AS animeId, anilist_id AS anilistId FROM imm_anime ORDER BY anime_id',
+      )
+      .all() as Array<{ animeId: number; anilistId: number | null }>;
+    assert.deepEqual(assignments, [
+      { animeId: 1, anilistId: 163132 },
+      { animeId: 2, anilistId: null },
+    ]);
+    assert.equal(videoAnimeId(db, 1), 1);
+    assert.equal(videoAnimeId(db, 2), 2);
+  });
+});
+
+test('automatic AniList update with unknown match confidence validates stored titles', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'actual show',
+      title: 'Actual Show',
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+    });
+    insertAnime(db, { animeId: 2, key: 'unrelated release', title: 'Unrelated Release' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    updateAnimeAnilistInfo(db, 2, {
+      anilistId: 163132,
+      titleRomaji: 'Actual Show',
+      titleEnglish: null,
+      titleNative: null,
+      episodesTotal: 12,
+    });
+
+    assert.deepEqual(animeIds(db), [1, 2]);
+    assert.equal(videoAnimeId(db, 2), 2);
+    assert.deepEqual(getAnimeMergeRecommendations(db), [{ recommendationId: 1, animeIds: [1, 2] }]);
+  });
+});
+
+test('stored AniList titles ignore season suffixes when validating an automatic merge', () => {
+  withDb((db) => {
+    insertAnime(db, {
+      animeId: 1,
+      key: 'legacy show',
+      title: 'Show Season 1',
+      anilistId: 163132,
+      titleRomaji: 'Show Season 1',
+    });
+    insertAnime(db, { animeId: 2, key: 'show season 1', title: 'Show Season 1' });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 1 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132);
+
+    assert.equal(summary.deletedAnimeRows, 1);
+    assert.deepEqual(animeIds(db), [1]);
+    assert.equal(videoAnimeId(db, 2), 1);
+    assert.deepEqual(getAnimeMergeRecommendations(db), []);
+  });
+});
+
+test('resolveAnimeAnilistConflict leaves an entry that already links elsewhere alone', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', anilistId: 163132 });
+    insertAnime(db, { animeId: 2, key: 'show s2', title: 'Show Season 2', anilistId: 999 });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 2 });
+
+    const summary = resolveAnimeAnilistConflict(db, 2, 163132);
+
+    assert.equal(videoAnimeId(db, 2), 2);
+    assert.ok(animeIds(db).includes(2));
+    assert.equal(
+      (
+        db.prepare('SELECT anilist_id AS anilistId FROM imm_anime WHERE anime_id = 2').get() as {
+          anilistId: number;
+        }
+      ).anilistId,
+      999,
+    );
+    assert.equal(summary.repaired, 0);
+    assert.equal(summary.movedVideos, 0);
+    assert.deepEqual(getAnimeMergeRecommendations(db), []);
+  });
+});
+
+test('automatic AniList update onto an entry that already links elsewhere does not throw', () => {
+  withDb((db) => {
+    insertAnime(db, { animeId: 1, key: 'show', title: 'Show', anilistId: 163132 });
+    insertAnime(db, { animeId: 2, key: 'show s2', title: 'Show Season 2', anilistId: 999 });
+    insertEpisode(db, { videoId: 1, animeId: 1, season: 1 });
+    insertEpisode(db, { videoId: 2, animeId: 2, season: 2 });
+
+    // Entry 2 explicitly links to 999; a later video re-resolving to entry 1's
+    // id must be refused, not written over the UNIQUE anilist_id column.
+    updateAnimeAnilistInfo(db, 2, {
+      anilistId: 163132,
+      titleRomaji: 'Show',
+      titleEnglish: null,
+      titleNative: null,
+      episodesTotal: 12,
+      exactTitleMatch: true,
+    });
+
+    assert.deepEqual(animeIds(db), [1, 2]);
+    assert.equal(
+      (
+        db.prepare('SELECT anilist_id AS anilistId FROM imm_anime WHERE anime_id = 2').get() as {
+          anilistId: number;
+        }
+      ).anilistId,
+      999,
+    );
+  });
+});

src/core/services/immersion-tracker/__tests__/query-split-modules.test.ts

@@ -50,6 +50,7 @@ import {
   updateAnimeAnilistInfo,
   upsertCoverArt,
 } from '../query-maintenance.js';
+import { deleteMaintenanceBatch } from '../query-delete-maintenance.js';
 import { getLocalEpochDay } from '../query-shared.js';
 import { EVENT_CARD_MINED, EVENT_SUBTITLE_LINE, SOURCE_TYPE_LOCAL } from '../types.js';
 
@@ -985,3 +986,197 @@ test('split maintenance helpers delete multiple sessions and whole videos with d
     cleanupDbPath(dbPath);
   }
 });
+
+test('delete maintenance batch preserves retained data across overlapping session, video, and anime targets', () => {
+  const { db, dbPath, stmts } = createDb();
+
+  try {
+    const retainedAnimeId = getOrCreateAnimeRecord(db, {
+      parsedTitle: 'Retained Anime',
+      canonicalTitle: 'Retained Anime',
+      anilistId: null,
+      titleRomaji: null,
+      titleEnglish: null,
+      titleNative: null,
+      metadataJson: null,
+    });
+    const deletedAnimeId = getOrCreateAnimeRecord(db, {
+      parsedTitle: 'Deleted Anime',
+      canonicalTitle: 'Deleted Anime',
+      anilistId: null,
+      titleRomaji: null,
+      titleEnglish: null,
+      titleNative: null,
+      metadataJson: null,
+    });
+    const retainedVideoId = getOrCreateVideoRecord(db, 'local:/tmp/batch-retain.mkv', {
+      canonicalTitle: 'Batch Retain',
+      sourcePath: '/tmp/batch-retain.mkv',
+      sourceUrl: null,
+      sourceType: SOURCE_TYPE_LOCAL,
+    });
+    const deletedVideoId = getOrCreateVideoRecord(db, 'local:/tmp/batch-video.mkv', {
+      canonicalTitle: 'Batch Video',
+      sourcePath: '/tmp/batch-video.mkv',
+      sourceUrl: null,
+      sourceType: SOURCE_TYPE_LOCAL,
+    });
+    const animeVideoId = getOrCreateVideoRecord(db, 'local:/tmp/batch-anime.mkv', {
+      canonicalTitle: 'Batch Anime',
+      sourcePath: '/tmp/batch-anime.mkv',
+      sourceUrl: null,
+      sourceType: SOURCE_TYPE_LOCAL,
+    });
+    for (const [videoId, animeId, episode] of [
+      [retainedVideoId, retainedAnimeId, 1],
+      [deletedVideoId, retainedAnimeId, 2],
+      [animeVideoId, deletedAnimeId, 1],
+    ] as const) {
+      linkVideoToAnimeRecord(db, videoId, {
+        animeId,
+        parsedBasename: `batch-${episode}.mkv`,
+        parsedTitle: animeId === retainedAnimeId ? 'Retained Anime' : 'Deleted Anime',
+        parsedSeason: 1,
+        parsedEpisode: episode,
+        parserSource: 'test',
+        parserConfidence: 1,
+        parseMetadataJson: null,
+      });
+    }
+
+    const startedAtMs = 1_700_000_000_000;
+    const deletedSessionId = startSessionRecord(db, retainedVideoId, startedAtMs).sessionId;
+    const retainedSessionId = startSessionRecord(
+      db,
+      retainedVideoId,
+      startedAtMs + 1_000,
+    ).sessionId;
+    const videoSessionId = startSessionRecord(db, deletedVideoId, startedAtMs + 2_000).sessionId;
+    const animeSessionId = startSessionRecord(db, animeVideoId, startedAtMs + 3_000).sessionId;
+    for (const [sessionId, sessionStartedAtMs] of [
+      [deletedSessionId, startedAtMs],
+      [retainedSessionId, startedAtMs + 1_000],
+      [videoSessionId, startedAtMs + 2_000],
+      [animeSessionId, startedAtMs + 3_000],
+    ] as const) {
+      finalizeSessionMetrics(db, sessionId, sessionStartedAtMs);
+    }
+
+    for (const [index, sessionId, videoId, animeId] of [
+      [1, deletedSessionId, retainedVideoId, retainedAnimeId],
+      [2, retainedSessionId, retainedVideoId, retainedAnimeId],
+      [3, videoSessionId, deletedVideoId, retainedAnimeId],
+      [4, animeSessionId, animeVideoId, deletedAnimeId],
+    ] as const) {
+      insertWordOccurrence(db, stmts, {
+        sessionId,
+        videoId,
+        animeId,
+        lineIndex: index,
+        text: '猫日',
+        word: { headword: '猫', word: '猫', reading: 'ねこ' },
+      });
+      insertKanjiOccurrence(db, stmts, {
+        sessionId,
+        videoId,
+        animeId,
+        lineIndex: index + 10,
+        text: '猫日',
+        kanji: '日',
+      });
+    }
+
+    const rollupDay = getLocalEpochDay(db, startedAtMs);
+    const rollupMonth = (
+      db
+        .prepare(
+          `SELECT CAST(strftime('%Y%m', CAST(? AS REAL) / 1000, 'unixepoch', 'localtime') AS INTEGER) AS rollupMonth`,
+        )
+        .get(startedAtMs) as { rollupMonth: number }
+    ).rollupMonth;
+    for (const videoId of [retainedVideoId, deletedVideoId, animeVideoId]) {
+      db.prepare(
+        `INSERT INTO imm_daily_rollups (
+          rollup_day, video_id, total_sessions, total_active_min, total_lines_seen,
+          total_tokens_seen, total_cards, CREATED_DATE, LAST_UPDATE_DATE
+        ) VALUES (?, ?, 99, 99, 99, 99, 99, ?, ?)`,
+      ).run(rollupDay, videoId, startedAtMs, startedAtMs);
+      db.prepare(
+        `INSERT INTO imm_monthly_rollups (
+          rollup_month, video_id, total_sessions, total_active_min, total_lines_seen,
+          total_tokens_seen, total_cards, CREATED_DATE, LAST_UPDATE_DATE
+        ) VALUES (?, ?, 99, 99, 99, 99, 99, ?, ?)`,
+      ).run(rollupMonth, videoId, startedAtMs, startedAtMs);
+    }
+
+    deleteMaintenanceBatch(db, [
+      { kind: 'session', sessionId: deletedSessionId },
+      { kind: 'session', sessionId: videoSessionId },
+      { kind: 'video', videoId: deletedVideoId },
+      { kind: 'video', videoId: animeVideoId },
+      { kind: 'anime', animeId: deletedAnimeId },
+    ]);
+
+    assert.deepEqual(db.prepare('SELECT session_id FROM imm_sessions').all(), [
+      { session_id: retainedSessionId },
+    ]);
+    assert.deepEqual(db.prepare('SELECT video_id FROM imm_videos').all(), [
+      { video_id: retainedVideoId },
+    ]);
+    assert.deepEqual(db.prepare('SELECT anime_id FROM imm_anime').all(), [
+      { anime_id: retainedAnimeId },
+    ]);
+    assert.equal(
+      (
+        db.prepare(`SELECT frequency FROM imm_words WHERE headword = '猫'`).get() as {
+          frequency: number;
+        }
+      ).frequency,
+      1,
+    );
+    assert.equal(
+      (
+        db.prepare(`SELECT frequency FROM imm_kanji WHERE kanji = '日'`).get() as {
+          frequency: number;
+        }
+      ).frequency,
+      1,
+    );
+    assert.deepEqual(
+      db.prepare('SELECT video_id, total_sessions FROM imm_daily_rollups').all() as Array<{
+        video_id: number;
+        total_sessions: number;
+      }>,
+      [{ video_id: retainedVideoId, total_sessions: 1 }],
+    );
+    assert.deepEqual(
+      db.prepare('SELECT video_id, total_sessions FROM imm_monthly_rollups').all() as Array<{
+        video_id: number;
+        total_sessions: number;
+      }>,
+      [{ video_id: retainedVideoId, total_sessions: 1 }],
+    );
+  } finally {
+    db.close();
+    cleanupDbPath(dbPath);
+  }
+});
+
+test('delete maintenance batch chunks id lists below the SQLite variable limit', () => {
+  const { db, dbPath } = createDb();
+
+  try {
+    const ids = Array.from({ length: 32_767 }, (_, index) => index + 1);
+
+    assert.doesNotThrow(() => {
+      deleteMaintenanceBatch(db, [
+        { kind: 'sessions', sessionIds: ids },
+        ...ids.map((videoId) => ({ kind: 'video' as const, videoId })),
+        ...ids.map((animeId) => ({ kind: 'anime' as const, animeId })),
+      ]);
+    });
+  } finally {
+    db.close();
+    cleanupDbPath(dbPath);
+  }
+});

src/core/services/immersion-tracker/anime-merge-recommendations.ts

@@ -0,0 +1,169 @@
+import type { DatabaseSync } from './sqlite';
+import { animeSeasonsAreMergeCompatible, getParsedSeasonsForAnime } from './anime-merge';
+import { toDbTimestamp } from './query-shared';
+import { normalizeAnimeIdentityKey } from './storage';
+import { nowMs } from './time';
+
+export interface AnimeMergeRecommendation {
+  recommendationId: number;
+  animeIds: [number, number];
+}
+
+export interface AnimeConflictRecommendationOptions {
+  survivor?: 'target' | 'existing';
+  /** Automatic matches must be exact; manual assignment is authoritative. */
+  matchConfidence?: 'exact' | 'weak' | 'manual';
+}
+
+interface AnimeTitleRow {
+  canonical_title: string;
+  title_romaji: string | null;
+  title_english: string | null;
+  title_native: string | null;
+}
+
+function getAnimeTitles(db: DatabaseSync, animeId: number): AnimeTitleRow | null {
+  return db
+    .prepare(
+      `SELECT canonical_title, title_romaji, title_english, title_native
+       FROM imm_anime
+       WHERE anime_id = ?`,
+    )
+    .get(animeId) as AnimeTitleRow | null;
+}
+
+function getParsedTitles(db: DatabaseSync, animeId: number): Array<string | null> {
+  return (
+    db.prepare('SELECT parsed_title FROM imm_videos WHERE anime_id = ?').all(animeId) as Array<{
+      parsed_title: string | null;
+    }>
+  ).map((row) => row.parsed_title);
+}
+
+function stripSeasonIdentitySuffix(title: string): string {
+  return title
+    .replace(/\bseason\s*\d{1,2}\b/gi, ' ')
+    .replace(/\b\d{1,2}(?:st|nd|rd|th)\s+season\b/gi, ' ')
+    .replace(/\bs\d{1,2}\b/gi, ' ');
+}
+
+export function hasExactStoredTitleMatch(
+  db: DatabaseSync,
+  targetAnimeId: number,
+  conflictAnimeId: number,
+): boolean {
+  const target = getAnimeTitles(db, targetAnimeId);
+  const conflict = getAnimeTitles(db, conflictAnimeId);
+  if (!target || !conflict) return false;
+  const targetKeys = [target.canonical_title, ...getParsedTitles(db, targetAnimeId)]
+    .filter((title): title is string => Boolean(title?.trim()))
+    .map((title) => normalizeAnimeIdentityKey(stripSeasonIdentitySuffix(title)))
+    .filter(Boolean);
+  const anilistTitleKeys = [
+    conflict.title_romaji,
+    conflict.title_english,
+    conflict.title_native,
+    conflict.canonical_title,
+  ]
+    .filter((title): title is string => Boolean(title?.trim()))
+    .map((title) => normalizeAnimeIdentityKey(stripSeasonIdentitySuffix(title)))
+    .filter(Boolean);
+  return targetKeys.some((key) => anilistTitleKeys.includes(key));
+}
+
+export function shouldRecommendAnilistConflict(
+  db: DatabaseSync,
+  targetAnimeId: number,
+  conflictAnimeId: number,
+  options: AnimeConflictRecommendationOptions,
+): boolean {
+  if (options.survivor === 'target' || options.matchConfidence === 'manual') return false;
+  if (
+    !animeSeasonsAreMergeCompatible(
+      getParsedSeasonsForAnime(db, targetAnimeId),
+      getParsedSeasonsForAnime(db, conflictAnimeId),
+    )
+  ) {
+    return false;
+  }
+  return (
+    options.matchConfidence === 'weak' ||
+    (options.matchConfidence === undefined &&
+      !hasExactStoredTitleMatch(db, targetAnimeId, conflictAnimeId))
+  );
+}
+
+export function recordAnimeMergeRecommendation(
+  db: DatabaseSync,
+  firstCandidateAnimeId: number,
+  secondCandidateAnimeId: number,
+  anilistId: number,
+): void {
+  const firstAnimeId = Math.min(firstCandidateAnimeId, secondCandidateAnimeId);
+  const secondAnimeId = Math.max(firstCandidateAnimeId, secondCandidateAnimeId);
+  const timestamp = toDbTimestamp(nowMs());
+  db.prepare(
+    `INSERT INTO imm_anime_merge_recommendations(
+       first_anime_id, second_anime_id, anilist_id, status, CREATED_DATE, LAST_UPDATE_DATE
+     ) VALUES (?, ?, ?, 'pending', ?, ?)
+     ON CONFLICT(first_anime_id, second_anime_id, anilist_id) DO UPDATE SET
+       LAST_UPDATE_DATE = excluded.LAST_UPDATE_DATE`,
+  ).run(firstAnimeId, secondAnimeId, anilistId, timestamp, timestamp);
+}
+
+export function hasDismissedAnimeMergeRecommendation(
+  db: DatabaseSync,
+  firstCandidateAnimeId: number,
+  secondCandidateAnimeId: number,
+): boolean {
+  const firstAnimeId = Math.min(firstCandidateAnimeId, secondCandidateAnimeId);
+  const secondAnimeId = Math.max(firstCandidateAnimeId, secondCandidateAnimeId);
+  return Boolean(
+    db
+      .prepare(
+        `SELECT 1
+         FROM imm_anime_merge_recommendations
+         WHERE first_anime_id = ?
+           AND second_anime_id = ?
+           AND status = 'dismissed'
+         LIMIT 1`,
+      )
+      .get(firstAnimeId, secondAnimeId),
+  );
+}
+
+export function getAnimeMergeRecommendations(db: DatabaseSync): AnimeMergeRecommendation[] {
+  return (
+    db
+      .prepare(
+        `SELECT recommendation_id AS recommendationId,
+                first_anime_id AS firstAnimeId,
+                second_anime_id AS secondAnimeId
+         FROM imm_anime_merge_recommendations
+         WHERE status = 'pending'
+         ORDER BY recommendation_id ASC`,
+      )
+      .all() as Array<{
+      recommendationId: number;
+      firstAnimeId: number;
+      secondAnimeId: number;
+    }>
+  ).map((row) => ({
+    recommendationId: row.recommendationId,
+    animeIds: [row.firstAnimeId, row.secondAnimeId],
+  }));
+}
+
+export function dismissAnimeMergeRecommendation(
+  db: DatabaseSync,
+  recommendationId: number,
+): boolean {
+  const result = db
+    .prepare(
+      `UPDATE imm_anime_merge_recommendations
+       SET status = 'dismissed', LAST_UPDATE_DATE = ?
+       WHERE recommendation_id = ? AND status = 'pending'`,
+    )
+    .run(toDbTimestamp(nowMs()), recommendationId) as { changes: number };
+  return result.changes > 0;
+}

src/core/services/immersion-tracker/anime-merge.ts

@@ -0,0 +1,296 @@
+import type { DatabaseSync } from './sqlite';
+import { recomputeLifetimeAnimeAggregatesInTransaction } from './lifetime';
+import { toDbTimestamp } from './query-shared';
+import { nowMs } from './time';
+
+/** Thrown when a move names an episode or destination entry that is not there. */
+export const UNKNOWN_MOVE_TARGET_MESSAGE = 'Unknown episode or target library entry';
+
+export interface AnimeMergeSummary {
+  /** Library entry that owns every moved episode once the merge finishes. */
+  survivingAnimeId: number;
+  /** Entries that were folded into the survivor and deleted. */
+  mergedAnimeIds: number[];
+  movedVideos: number;
+}
+
+export interface VideoMoveSummary {
+  targetAnimeId: number;
+  /** Previous owner, or null when the episode had no library entry yet. */
+  previousAnimeId: number | null;
+  /** True when the previous owner was left empty and pruned. */
+  removedPreviousAnime: boolean;
+}
+
+interface AnimeMetadataRow {
+  normalized_title_key: string;
+  anilist_id: number | null;
+  title_romaji: string | null;
+  title_english: string | null;
+  title_native: string | null;
+  episodes_total: number | null;
+  description: string | null;
+}
+
+function emptyMergeSummary(survivingAnimeId: number): AnimeMergeSummary {
+  return { survivingAnimeId, mergedAnimeIds: [], movedVideos: 0 };
+}
+
+function runInTransaction<T>(db: DatabaseSync, work: () => T): T {
+  db.exec('BEGIN IMMEDIATE');
+  try {
+    const result = work();
+    db.exec('COMMIT');
+    return result;
+  } catch (error) {
+    db.exec('ROLLBACK');
+    throw error;
+  }
+}
+
+function readAnimeMetadata(db: DatabaseSync, animeId: number): AnimeMetadataRow | null {
+  return (db
+    .prepare(
+      `
+        SELECT normalized_title_key, anilist_id, title_romaji, title_english, title_native, episodes_total, description
+        FROM imm_anime
+        WHERE anime_id = ?
+      `,
+    )
+    .get(animeId) ?? null) as AnimeMetadataRow | null;
+}
+
+function animeExists(db: DatabaseSync, animeId: number): boolean {
+  return Boolean(db.prepare('SELECT 1 FROM imm_anime WHERE anime_id = ?').get(animeId));
+}
+
+function hasAnimeReferences(db: DatabaseSync, animeId: number): boolean {
+  const row = db
+    .prepare(
+      `
+        SELECT 1 AS found
+        WHERE EXISTS (SELECT 1 FROM imm_videos WHERE anime_id = ?)
+           OR EXISTS (SELECT 1 FROM imm_subtitle_lines WHERE anime_id = ?)
+      `,
+    )
+    .get(animeId, animeId) as { found: number } | null;
+  return Boolean(row);
+}
+
+/**
+ * Distinct explicit seasons behind a library entry. Videos with no parsed
+ * season are ignored, so an entry built from `Show - 03.mkv` style filenames
+ * reports an empty set rather than a bogus season.
+ */
+export function getParsedSeasonsForAnime(db: DatabaseSync, animeId: number): Set<number> {
+  const rows = db
+    .prepare(
+      `
+        SELECT DISTINCT parsed_season AS season
+        FROM imm_videos
+        WHERE anime_id = ?
+          AND parsed_season IS NOT NULL
+          AND parsed_season > 0
+      `,
+    )
+    .all(animeId) as Array<{ season: number }>;
+  return new Set(rows.map((row) => row.season));
+}
+
+/**
+ * Two entries are safe to fold together when neither spans more than one
+ * explicit season and they do not disagree about which season that is. A
+ * seasonless entry is compatible with anything single-season: those are the
+ * `Show - 03.mkv` vs `Show.S01E03.mkv` splits that produce duplicate cards.
+ */
+export function animeSeasonsAreMergeCompatible(a: Set<number>, b: Set<number>): boolean {
+  if (a.size > 1 || b.size > 1) return false;
+  if (a.size === 0 || b.size === 0) return true;
+  return [...a][0] === [...b][0];
+}
+
+/**
+ * Fill in whatever the target is missing from a source row that is on its way
+ * out. Must run after the source row is deleted: imm_anime.anilist_id is
+ * UNIQUE, so the two rows cannot hold the same id at once.
+ */
+function absorbAnimeMetadata(
+  db: DatabaseSync,
+  targetAnimeId: number,
+  source: AnimeMetadataRow | null,
+  updatedAt: string,
+): void {
+  if (!source) return;
+  db.prepare(
+    `
+      UPDATE imm_anime
+      SET
+        anilist_id = COALESCE(anilist_id, ?),
+        title_romaji = COALESCE(title_romaji, ?),
+        title_english = COALESCE(title_english, ?),
+        title_native = COALESCE(title_native, ?),
+        episodes_total = COALESCE(episodes_total, ?),
+        description = COALESCE(description, ?),
+        LAST_UPDATE_DATE = ?
+      WHERE anime_id = ?
+    `,
+  ).run(
+    source.anilist_id,
+    source.title_romaji,
+    source.title_english,
+    source.title_native,
+    source.episodes_total,
+    source.description,
+    updatedAt,
+    targetAnimeId,
+  );
+}
+
+/**
+ * Fold `sourceAnimeIds` into `targetAnimeId`: every episode and subtitle line
+ * is repointed, metadata the target is missing is inherited from the sources,
+ * and the emptied source rows are deleted.
+ *
+ * Assumes the caller already holds a write transaction and refreshes the
+ * per-anime lifetime aggregates afterwards; use {@link mergeAnimeRecords}
+ * otherwise.
+ */
+export function mergeAnimeRecordsInTransaction(
+  db: DatabaseSync,
+  targetAnimeId: number,
+  sourceAnimeIds: number[],
+): AnimeMergeSummary {
+  const summary = emptyMergeSummary(targetAnimeId);
+  if (!animeExists(db, targetAnimeId)) {
+    return summary;
+  }
+
+  const updatedAt = toDbTimestamp(nowMs());
+  const sourceVideosStmt = db.prepare(
+    'SELECT video_id AS videoId FROM imm_videos WHERE anime_id = ?',
+  );
+  const moveVideosStmt = db.prepare(
+    'UPDATE imm_videos SET anime_id = ?, LAST_UPDATE_DATE = ? WHERE anime_id = ?',
+  );
+  // Repointed per video rather than by anime_id: lines recorded before the
+  // async title parse assigns the link are stored with a NULL anime_id, and
+  // matching on the source id would strand them unattributed.
+  const moveLinesStmt = db.prepare(
+    'UPDATE imm_subtitle_lines SET anime_id = ?, LAST_UPDATE_DATE = ? WHERE video_id = ?',
+  );
+  const dropLifetimeStmt = db.prepare('DELETE FROM imm_lifetime_anime WHERE anime_id = ?');
+  const sourceAliasesStmt = db.prepare(
+    'SELECT normalized_title_key AS normalizedTitleKey FROM imm_anime_title_aliases WHERE anime_id = ?',
+  );
+  const upsertAliasStmt = db.prepare(
+    `INSERT INTO imm_anime_title_aliases(normalized_title_key, anime_id, CREATED_DATE, LAST_UPDATE_DATE)
+     VALUES (?, ?, ?, ?)
+     ON CONFLICT(normalized_title_key) DO UPDATE SET
+       anime_id = excluded.anime_id,
+       LAST_UPDATE_DATE = excluded.LAST_UPDATE_DATE`,
+  );
+  const dropSourceAliasesStmt = db.prepare(
+    'DELETE FROM imm_anime_title_aliases WHERE anime_id = ?',
+  );
+  const dropAnimeStmt = db.prepare('DELETE FROM imm_anime WHERE anime_id = ?');
+
+  for (const sourceAnimeId of new Set(sourceAnimeIds)) {
+    if (sourceAnimeId === targetAnimeId || !animeExists(db, sourceAnimeId)) {
+      continue;
+    }
+
+    const sourceMetadata = readAnimeMetadata(db, sourceAnimeId);
+    const sourceAliases = sourceAliasesStmt.all(sourceAnimeId) as Array<{
+      normalizedTitleKey: string;
+    }>;
+    const sourceVideoIds = (sourceVideosStmt.all(sourceAnimeId) as Array<{ videoId: number }>).map(
+      (row) => row.videoId,
+    );
+    const moved = moveVideosStmt.run(targetAnimeId, updatedAt, sourceAnimeId) as {
+      changes: number;
+    };
+    for (const videoId of sourceVideoIds) {
+      moveLinesStmt.run(targetAnimeId, updatedAt, videoId);
+    }
+    dropSourceAliasesStmt.run(sourceAnimeId);
+    for (const alias of [
+      ...(sourceMetadata ? [sourceMetadata.normalized_title_key] : []),
+      ...sourceAliases.map((row) => row.normalizedTitleKey),
+    ]) {
+      upsertAliasStmt.run(alias, targetAnimeId, updatedAt, updatedAt);
+    }
+    dropLifetimeStmt.run(sourceAnimeId);
+    dropAnimeStmt.run(sourceAnimeId);
+    absorbAnimeMetadata(db, targetAnimeId, sourceMetadata, updatedAt);
+
+    summary.mergedAnimeIds.push(sourceAnimeId);
+    summary.movedVideos += moved.changes;
+  }
+
+  return summary;
+}
+
+export function mergeAnimeRecords(
+  db: DatabaseSync,
+  targetAnimeId: number,
+  sourceAnimeIds: number[],
+): AnimeMergeSummary {
+  return runInTransaction(db, () => {
+    const summary = mergeAnimeRecordsInTransaction(db, targetAnimeId, sourceAnimeIds);
+    if (summary.mergedAnimeIds.length > 0) {
+      recomputeLifetimeAnimeAggregatesInTransaction(db);
+    }
+    return summary;
+  });
+}
+
+/**
+ * Move a single episode to another library entry, pruning the previous owner
+ * when it is left with nothing.
+ */
+export function moveVideoToAnime(
+  db: DatabaseSync,
+  videoId: number,
+  targetAnimeId: number,
+): VideoMoveSummary {
+  return runInTransaction(db, () => {
+    const videoRow = db
+      .prepare('SELECT anime_id AS animeId FROM imm_videos WHERE video_id = ?')
+      .get(videoId) as { animeId: number | null } | null;
+    if (!videoRow || !animeExists(db, targetAnimeId)) {
+      throw new Error(UNKNOWN_MOVE_TARGET_MESSAGE);
+    }
+
+    const previousAnimeId = videoRow.animeId;
+    if (previousAnimeId === targetAnimeId) {
+      db.prepare(
+        'UPDATE imm_videos SET anime_assignment_locked = 1, LAST_UPDATE_DATE = ? WHERE video_id = ?',
+      ).run(toDbTimestamp(nowMs()), videoId);
+      return { targetAnimeId, previousAnimeId, removedPreviousAnime: false };
+    }
+
+    const updatedAt = toDbTimestamp(nowMs());
+    db.prepare(
+      `UPDATE imm_videos
+       SET anime_id = ?, anime_assignment_locked = 1, LAST_UPDATE_DATE = ?
+       WHERE video_id = ?`,
+    ).run(targetAnimeId, updatedAt, videoId);
+    db.prepare(
+      'UPDATE imm_subtitle_lines SET anime_id = ?, LAST_UPDATE_DATE = ? WHERE video_id = ?',
+    ).run(targetAnimeId, updatedAt, videoId);
+
+    let removedPreviousAnime = false;
+    if (previousAnimeId !== null && !hasAnimeReferences(db, previousAnimeId)) {
+      // The emptied entry's metadata is deliberately dropped rather than
+      // absorbed. A move says "this episode belongs elsewhere", not "these are
+      // the same show", and the entry being emptied is usually a mis-parse
+      // whose AniList link would be wrong for the target.
+      db.prepare('DELETE FROM imm_lifetime_anime WHERE anime_id = ?').run(previousAnimeId);
+      db.prepare('DELETE FROM imm_anime WHERE anime_id = ?').run(previousAnimeId);
+      removedPreviousAnime = true;
+    }
+
+    recomputeLifetimeAnimeAggregatesInTransaction(db);
+    return { targetAnimeId, previousAnimeId, removedPreviousAnime };
+  });
+}

src/core/services/immersion-tracker/anime-season-repair.ts

@@ -1,4 +1,16 @@
 import type { DatabaseSync } from './sqlite';
+import {
+  animeSeasonsAreMergeCompatible,
+  getParsedSeasonsForAnime,
+  mergeAnimeRecordsInTransaction,
+} from './anime-merge';
+import {
+  hasExactStoredTitleMatch,
+  hasDismissedAnimeMergeRecommendation,
+  recordAnimeMergeRecommendation,
+  shouldRecommendAnilistConflict,
+  type AnimeConflictRecommendationOptions,
+} from './anime-merge-recommendations';
 import { getOrCreateAnimeRecord } from './storage';
 import { toDbTimestamp } from './query-shared';
 import { nowMs } from './time';
@@ -8,8 +20,33 @@ export interface AnimeSeasonRepairSummary {
   repaired: number;
   movedVideos: number;
   deletedAnimeRows: number;
+  /**
+   * Entry that owns the videos afterwards when two rows were folded together,
+   * so callers can keep pointing at a row that still exists.
+   */
+  survivingAnimeId: number | null;
+  /** True when an ambiguous AniList collision was saved for user review. */
+  mergeRecommended: boolean;
+  /** True when automatic metadata must not assign the colliding AniList id. */
+  anilistAssignmentBlocked: boolean;
 }
 
+export interface AnimeAnilistConflictOptions extends AnimeConflictRecommendationOptions {
+  /**
+   * Which row keeps its identity when two entries claim the same AniList id.
+   * `existing` (the default) keeps the row that already held the id, so
+   * automatic cover-art resolution does not rename a card under the user;
+   * `target` keeps the row the user is acting on.
+   */
+  survivor?: 'target' | 'existing';
+}
+
+export {
+  dismissAnimeMergeRecommendation,
+  getAnimeMergeRecommendations,
+  type AnimeMergeRecommendation,
+} from './anime-merge-recommendations';
+
 interface AnimeRow {
   anime_id: number;
   anilist_id: number | null;
@@ -24,6 +61,7 @@ interface ParsedVideoRow {
   video_id: number;
   parsed_title: string | null;
   parsed_season: number | null;
+  anime_assignment_locked: number;
 }
 
 interface RedistributeOptions {
@@ -38,6 +76,9 @@ function emptySummary(scanned = 0): AnimeSeasonRepairSummary {
     repaired: 0,
     movedVideos: 0,
     deletedAnimeRows: 0,
+    survivingAnimeId: null,
+    mergeRecommended: false,
+    anilistAssignmentBlocked: false,
   };
 }
 
@@ -49,6 +90,9 @@ function mergeSummary(
   target.repaired += source.repaired;
   target.movedVideos += source.movedVideos;
   target.deletedAnimeRows += source.deletedAnimeRows;
+  target.survivingAnimeId = source.survivingAnimeId ?? target.survivingAnimeId;
+  target.mergeRecommended ||= source.mergeRecommended;
+  target.anilistAssignmentBlocked ||= source.anilistAssignmentBlocked;
   return target;
 }
 
@@ -94,7 +138,7 @@ function getParsedVideos(db: DatabaseSync, animeId: number): ParsedVideoRow[] {
   return db
     .prepare(
       `
-        SELECT video_id, parsed_title, parsed_season
+        SELECT video_id, parsed_title, parsed_season, anime_assignment_locked
         FROM imm_videos
         WHERE anime_id = ?
         ORDER BY video_id ASC
@@ -188,6 +232,9 @@ function redistributeAnimeRowByParsedSeasonsInTransaction(
   const targetBySeason = new Map<number, number>();
 
   for (const video of videos) {
+    if (video.anime_assignment_locked === 1) {
+      continue;
+    }
     const parsedTitle = video.parsed_title?.trim();
     const season = normalizeSeason(video.parsed_season);
     if (!parsedTitle || season === null) {
@@ -301,10 +348,18 @@ export function repairLegacySeasonlessAnimeRows(db: DatabaseSync): AnimeSeasonRe
   });
 }
 
+/**
+ * Two library entries cannot both hold the same AniList id
+ * (`imm_anime.anilist_id` is UNIQUE). Fold an automatic collision only when
+ * exact title evidence and compatible parsed seasons make it safe. Persist a
+ * review recommendation for compatible weak matches. Fall back to legacy
+ * season redistribution when the conflicting row spans several seasons.
+ */
 export function resolveAnimeAnilistConflict(
   db: DatabaseSync,
   targetAnimeId: number,
   anilistId: number,
+  options: AnimeAnilistConflictOptions = {},
 ): AnimeSeasonRepairSummary {
   const conflict = db
     .prepare(
@@ -321,10 +376,100 @@ export function resolveAnimeAnilistConflict(
     return emptySummary();
   }
 
-  return runInTransaction(db, () =>
+  return runInTransaction(db, () => {
-    redistributeAnimeRowByParsedSeasonsInTransaction(db, conflict.animeId, {
+    const targetRow = getAnimeRow(db, targetAnimeId);
+    if (
+      options.survivor !== 'target' &&
+      targetRow?.anilist_id != null &&
+      targetRow.anilist_id !== anilistId
+    ) {
+      // An automatic lookup disagreeing with an existing explicit link is a
+      // mis-resolution, not evidence that either row should move or merge. The
+      // colliding id must not be assigned either: another row owns it and
+      // imm_anime.anilist_id is UNIQUE.
+      const summary = emptySummary(1);
+      summary.anilistAssignmentBlocked = true;
+      return summary;
+    }
+    const isManual = options.survivor === 'target' || options.matchConfidence === 'manual';
+    if (!isManual && hasDismissedAnimeMergeRecommendation(db, targetAnimeId, conflict.animeId)) {
+      const summary = emptySummary(1);
+      summary.anilistAssignmentBlocked = true;
+      return summary;
+    }
+    const targetSeasons = getParsedSeasonsForAnime(db, targetAnimeId);
+    const conflictSeasons = getParsedSeasonsForAnime(db, conflict.animeId);
+    if (
+      !isManual &&
+      targetSeasons.size === 1 &&
+      conflictSeasons.size === 1 &&
+      [...targetSeasons][0] !== [...conflictSeasons][0]
+    ) {
+      const summary = emptySummary(1);
+      summary.anilistAssignmentBlocked = true;
+      return summary;
+    }
+    if (canMergeAnilistConflict(db, targetAnimeId, conflict.animeId, anilistId, options)) {
+      const survivingAnimeId = options.survivor === 'target' ? targetAnimeId : conflict.animeId;
+      const absorbedAnimeId = survivingAnimeId === targetAnimeId ? conflict.animeId : targetAnimeId;
+      const merge = mergeAnimeRecordsInTransaction(db, survivingAnimeId, [absorbedAnimeId]);
+      const summary = emptySummary(1);
+      summary.movedVideos = merge.movedVideos;
+      summary.deletedAnimeRows = merge.mergedAnimeIds.length;
+      if (merge.mergedAnimeIds.length > 0) {
+        summary.repaired = 1;
+        // Only reported once a row really absorbed the other, so callers never
+        // follow this to an anime id that was never written.
+        summary.survivingAnimeId = survivingAnimeId;
+      }
+      // Lifetime summaries are rebuilt by the caller off this summary, the same
+      // as the redistribution path below.
+      return summary;
+    }
+
+    if (shouldRecommendAnilistConflict(db, targetAnimeId, conflict.animeId, options)) {
+      recordAnimeMergeRecommendation(db, targetAnimeId, conflict.animeId, anilistId);
+      const summary = emptySummary(1);
+      summary.mergeRecommended = true;
+      return summary;
+    }
+
+    return redistributeAnimeRowByParsedSeasonsInTransaction(db, conflict.animeId, {
       transferAnilistToAnimeId: targetAnimeId,
       overwriteTargetAnilist: true,
-    }),
+    });
+  });
+}
+
+function canMergeAnilistConflict(
+  db: DatabaseSync,
+  targetAnimeId: number,
+  conflictAnimeId: number,
+  anilistId: number,
+  options: AnimeAnilistConflictOptions,
+): boolean {
+  const targetRow = getAnimeRow(db, targetAnimeId);
+  if (!targetRow) {
+    // Nothing to merge with a row that no longer exists (a stale id from the
+    // caller); fall through to the redistribution path.
+    return false;
+  }
+  if (options.survivor !== 'target') {
+    // The target is the row about to disappear here, so an existing link of its
+    // own means this is a mis-resolution rather than a duplicate: leave it be.
+    if (targetRow.anilist_id != null && targetRow.anilist_id !== anilistId) {
+      return false;
+    }
+  }
+  if (
+    options.matchConfidence === 'weak' ||
+    (options.matchConfidence === undefined &&
+      !hasExactStoredTitleMatch(db, targetAnimeId, conflictAnimeId))
+  ) {
+    return false;
+  }
+  return animeSeasonsAreMergeCompatible(
+    getParsedSeasonsForAnime(db, targetAnimeId),
+    getParsedSeasonsForAnime(db, conflictAnimeId),
   );
 }

src/core/services/immersion-tracker/delete-maintenance-scheduler.test.ts

@@ -0,0 +1,160 @@
+import assert from 'node:assert/strict';
+import test from 'node:test';
+import { DeleteMaintenanceScheduler } from './delete-maintenance-scheduler';
+import type { DeleteMaintenanceTask } from './delete-maintenance';
+
+test('scheduler batches same-turn requests and balances busy state', async () => {
+  const tasks: DeleteMaintenanceTask[] = [];
+  const states: string[] = [];
+  const scheduler = new DeleteMaintenanceScheduler({
+    batchWindowMs: 0,
+    runTask: async (task) => {
+      tasks.push(task);
+    },
+    onBusy: () => states.push('busy'),
+    onIdle: () => states.push('idle'),
+  });
+
+  const first = scheduler.enqueue(() => ({ kind: 'session', sessionId: 1 }));
+  const second = scheduler.enqueue(() => ({ kind: 'sessions', sessionIds: [2, 3] }));
+  const third = scheduler.enqueue(() => null);
+  await Promise.all([first, second, third]);
+
+  assert.deepEqual(tasks, [
+    {
+      kind: 'batch',
+      tasks: [
+        { kind: 'session', sessionId: 1 },
+        { kind: 'sessions', sessionIds: [2, 3] },
+      ],
+    },
+  ]);
+  assert.deepEqual(states, ['busy', 'idle']);
+});
+
+test('scheduler rejects enqueue after destruction without entering busy state', async () => {
+  let busyCalls = 0;
+  let runCalls = 0;
+  const scheduler = new DeleteMaintenanceScheduler({
+    batchWindowMs: 0,
+    runTask: async () => {
+      runCalls += 1;
+    },
+    onBusy: () => {
+      busyCalls += 1;
+    },
+    onIdle: () => {},
+  });
+  scheduler.destroy();
+
+  await assert.rejects(
+    scheduler.enqueue(() => ({ kind: 'session', sessionId: 1 })),
+    /shutting down/,
+  );
+  assert.equal(busyCalls, 0);
+  assert.equal(runCalls, 0);
+});
+
+test('scheduler rejects every request in a batch when the maintenance task fails', async () => {
+  const failure = new Error('maintenance failed');
+  const scheduler = new DeleteMaintenanceScheduler({
+    batchWindowMs: 0,
+    runTask: async () => {
+      throw failure;
+    },
+    onBusy: () => {},
+    onIdle: () => {},
+  });
+
+  const first = scheduler.enqueue(() => ({ kind: 'session', sessionId: 1 }));
+  const second = scheduler.enqueue(() => ({ kind: 'session', sessionId: 2 }));
+
+  const results = await Promise.allSettled([first, second]);
+  assert.deepEqual(
+    results.map((result) => (result.status === 'rejected' ? result.reason : null)),
+    [failure, failure],
+  );
+});
+
+test('scheduler rejects only the request whose task resolution fails', async () => {
+  const failure = new Error('resolution failed');
+  const tasks: DeleteMaintenanceTask[] = [];
+  const scheduler = new DeleteMaintenanceScheduler({
+    batchWindowMs: 0,
+    runTask: async (task) => {
+      tasks.push(task);
+    },
+    onBusy: () => {},
+    onIdle: () => {},
+  });
+
+  const failed = scheduler.enqueue(() => {
+    throw failure;
+  });
+  const succeeded = scheduler.enqueue(() => ({ kind: 'session', sessionId: 2 }));
+
+  const results = await Promise.allSettled([failed, succeeded]);
+  assert.equal(results[0]?.status, 'rejected');
+  assert.equal(results[0]?.status === 'rejected' ? results[0].reason : null, failure);
+  assert.equal(results[1]?.status, 'fulfilled');
+  assert.deepEqual(tasks, [{ kind: 'session', sessionId: 2 }]);
+});
+
+test('scheduler does not schedule another drain when the queue is empty', async () => {
+  const originalSetTimeout = globalThis.setTimeout;
+  let timerCalls = 0;
+  globalThis.setTimeout = ((handler: TimerHandler, timeout?: number, ...args: unknown[]) => {
+    timerCalls += 1;
+    return originalSetTimeout(handler, timeout, ...args);
+  }) as typeof setTimeout;
+
+  try {
+    const scheduler = new DeleteMaintenanceScheduler({
+      batchWindowMs: 0,
+      runTask: async () => {},
+      onBusy: () => {},
+      onIdle: () => {},
+    });
+
+    await scheduler.enqueue(() => ({ kind: 'session', sessionId: 1 }));
+    assert.equal(timerCalls, 1);
+  } finally {
+    globalThis.setTimeout = originalSetTimeout;
+  }
+});
+
+test('scheduler serializes batches and rejects requests queued at destruction', async () => {
+  const releases: Array<() => void> = [];
+  let activeTasks = 0;
+  let maxActiveTasks = 0;
+  const scheduler = new DeleteMaintenanceScheduler({
+    batchWindowMs: 0,
+    runTask: async () => {
+      activeTasks += 1;
+      maxActiveTasks = Math.max(maxActiveTasks, activeTasks);
+      await new Promise<void>((resolve) => releases.push(resolve));
+      activeTasks -= 1;
+    },
+    onBusy: () => {},
+    onIdle: () => {},
+  });
+
+  const first = scheduler.enqueue(() => ({ kind: 'session', sessionId: 1 }));
+  const maxPollAttempts = 100;
+  let pollAttempts = 0;
+  while (releases.length === 0 && pollAttempts < maxPollAttempts) {
+    pollAttempts += 1;
+    await new Promise<void>((resolve) => setTimeout(resolve, 0));
+  }
+  assert.ok(
+    releases.length > 0,
+    `runTask did not produce a release after ${maxPollAttempts} polling attempts`,
+  );
+  const queued = scheduler.enqueue(() => ({ kind: 'session', sessionId: 2 }));
+  scheduler.destroy();
+
+  await assert.rejects(queued, /shutting down/);
+  releases[0]?.();
+  await first;
+  assert.equal(maxActiveTasks, 1);
+});

src/core/services/immersion-tracker/delete-maintenance-scheduler.ts

@@ -0,0 +1,105 @@
+import type { DeleteMaintenanceOperation, DeleteMaintenanceTask } from './delete-maintenance';
+
+type ResolveDeleteMaintenanceOperation = () =>
+  | DeleteMaintenanceOperation
+  | null
+  | Promise<DeleteMaintenanceOperation | null>;
+
+interface PendingDeleteMaintenanceRequest {
+  resolveTask: ResolveDeleteMaintenanceOperation;
+  resolve: () => void;
+  reject: (error: unknown) => void;
+}
+
+interface DeleteMaintenanceSchedulerOptions {
+  batchWindowMs: number;
+  runTask: (task: DeleteMaintenanceTask) => Promise<void>;
+  onBusy: () => void;
+  onIdle: () => void;
+}
+
+export class DeleteMaintenanceScheduler {
+  private readonly pendingRequests: PendingDeleteMaintenanceRequest[] = [];
+  private running = false;
+  private drainTimer: ReturnType<typeof setTimeout> | null = null;
+  private pendingTaskCount = 0;
+  private destroyed = false;
+
+  constructor(private readonly options: DeleteMaintenanceSchedulerOptions) {}
+
+  enqueue(resolveTask: ResolveDeleteMaintenanceOperation): Promise<void> {
+    if (this.destroyed) {
+      return Promise.reject(new Error('Immersion tracker is shutting down'));
+    }
+
+    if (this.pendingTaskCount === 0) this.options.onBusy();
+    this.pendingTaskCount += 1;
+
+    const result = new Promise<void>((resolve, reject) => {
+      this.pendingRequests.push({ resolveTask, resolve, reject });
+      this.scheduleDrain();
+    });
+
+    return result.finally(() => {
+      this.pendingTaskCount -= 1;
+      if (this.pendingTaskCount === 0) this.options.onIdle();
+    });
+  }
+
+  destroy(): void {
+    if (this.destroyed) return;
+    this.destroyed = true;
+    if (this.drainTimer) {
+      clearTimeout(this.drainTimer);
+      this.drainTimer = null;
+    }
+    const error = new Error('Immersion tracker is shutting down');
+    for (const request of this.pendingRequests.splice(0)) request.reject(error);
+  }
+
+  private scheduleDrain(): void {
+    if (this.destroyed || this.running || this.drainTimer || this.pendingRequests.length === 0) {
+      return;
+    }
+    this.drainTimer = setTimeout(() => {
+      this.drainTimer = null;
+      void this.drain();
+    }, this.options.batchWindowMs);
+  }
+
+  private async drain(): Promise<void> {
+    if (this.running || this.pendingRequests.length === 0) return;
+    this.running = true;
+    const requests = this.pendingRequests.splice(0);
+    const runnable: Array<{
+      request: PendingDeleteMaintenanceRequest;
+      task: DeleteMaintenanceOperation;
+    }> = [];
+
+    for (const request of requests) {
+      try {
+        const task = await request.resolveTask();
+        if (task) runnable.push({ request, task });
+        else request.resolve();
+      } catch (error) {
+        request.reject(error);
+      }
+    }
+
+    if (runnable.length > 0) {
+      const task: DeleteMaintenanceTask =
+        runnable.length === 1
+          ? runnable[0]!.task
+          : { kind: 'batch', tasks: runnable.map((entry) => entry.task) };
+      try {
+        await this.options.runTask(task);
+        for (const { request } of runnable) request.resolve();
+      } catch (error) {
+        for (const { request } of runnable) request.reject(error);
+      }
+    }
+
+    this.running = false;
+    this.scheduleDrain();
+  }
+}

src/core/services/immersion-tracker/delete-maintenance-worker-runtime.test.ts

@@ -0,0 +1,239 @@
+import assert from 'node:assert/strict';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import test from 'node:test';
+import {
+  DeleteMaintenanceWorkerRuntime,
+  resolveDeleteMaintenanceWorkerPath,
+} from './delete-maintenance-worker-runtime';
+import { executeDeleteMaintenanceTask } from './delete-maintenance';
+import { startSessionRecord } from './session';
+import { Database } from './sqlite';
+import { applyPragmas, ensureSchema, getOrCreateVideoRecord } from './storage';
+
+type FakeWorkerListener = (value: never) => void;
+
+function createFakeWorker() {
+  const listeners = new Map<string, FakeWorkerListener>();
+  const terminationState = { calls: 0 };
+  const worker = {
+    once(event: string, listener: FakeWorkerListener) {
+      listeners.set(event, listener);
+      return this;
+    },
+    terminate: async () => {
+      terminationState.calls += 1;
+      return 0;
+    },
+  };
+  return { worker, listeners, terminationState };
+}
+
+type FakeWorker = ReturnType<typeof createFakeWorker>['worker'];
+
+test('a delete batch rebuilds lifetime summaries once', () => {
+  const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'subminer-delete-batch-test-'));
+  const dbPath = path.join(tempDir, 'immersion.sqlite');
+  let db = new Database(dbPath);
+
+  try {
+    applyPragmas(db);
+    ensureSchema(db);
+    const videoId = getOrCreateVideoRecord(db, 'local:/tmp/batch-delete.mkv', {
+      canonicalTitle: 'Batch Delete',
+      sourcePath: '/tmp/batch-delete.mkv',
+      sourceUrl: null,
+      sourceType: 1,
+    });
+    const firstSessionId = startSessionRecord(db, videoId, 1_000).sessionId;
+    const secondSessionId = startSessionRecord(db, videoId, 2_000).sessionId;
+    const deletedVideoId = getOrCreateVideoRecord(db, 'local:/tmp/batch-delete-video.mkv', {
+      canonicalTitle: 'Batch Delete Video',
+      sourcePath: '/tmp/batch-delete-video.mkv',
+      sourceUrl: null,
+      sourceType: 1,
+    });
+    startSessionRecord(db, deletedVideoId, 3_000);
+    db.exec(`
+      CREATE TABLE delete_rebuild_audit (id INTEGER PRIMARY KEY);
+      CREATE TRIGGER count_delete_lifetime_rebuild
+      AFTER UPDATE OF last_rebuilt_ms ON imm_lifetime_global
+      BEGIN
+        INSERT INTO delete_rebuild_audit (id) VALUES (NULL);
+      END;
+    `);
+    db.close();
+
+    executeDeleteMaintenanceTask(dbPath, {
+      kind: 'batch',
+      tasks: [
+        { kind: 'session', sessionId: firstSessionId },
+        { kind: 'video', videoId: deletedVideoId },
+      ],
+    });
+
+    db = new Database(dbPath);
+    const audit = db.prepare('SELECT COUNT(*) AS total FROM delete_rebuild_audit').get() as {
+      total: number;
+    };
+    const retainedSession = db
+      .prepare('SELECT session_id AS sessionId FROM imm_sessions WHERE video_id = ?')
+      .get(videoId) as { sessionId: number } | null;
+    const deletedVideo = db
+      .prepare('SELECT video_id AS videoId FROM imm_videos WHERE video_id = ?')
+      .get(deletedVideoId) as { videoId: number } | null;
+    assert.equal(retainedSession?.sessionId, secondSessionId);
+    assert.equal(deletedVideo, undefined);
+    assert.equal(
+      audit.total,
+      2,
+      'one rebuild performs exactly its reset and final global summary writes',
+    );
+  } finally {
+    try {
+      db.close();
+    } catch {
+      // The setup connection closes before maintenance runs.
+    }
+    fs.rmSync(tempDir, { recursive: true, force: true });
+  }
+});
+
+test(
+  'compiled delete worker removes data through its separate database connection',
+  { skip: resolveDeleteMaintenanceWorkerPath() === null },
+  async () => {
+    const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'subminer-delete-worker-test-'));
+    const dbPath = path.join(tempDir, 'immersion.sqlite');
+    const runtime = new DeleteMaintenanceWorkerRuntime();
+    let db = new Database(dbPath);
+
+    try {
+      applyPragmas(db);
+      ensureSchema(db);
+      const videoId = getOrCreateVideoRecord(db, 'local:/tmp/worker-delete.mkv', {
+        canonicalTitle: 'Worker Delete',
+        sourcePath: '/tmp/worker-delete.mkv',
+        sourceUrl: null,
+        sourceType: 1,
+      });
+      const firstSessionId = startSessionRecord(db, videoId, 1_000).sessionId;
+      const secondSessionId = startSessionRecord(db, videoId, 2_000).sessionId;
+      db.close();
+
+      await runtime.run(dbPath, {
+        kind: 'batch',
+        tasks: [
+          { kind: 'session', sessionId: firstSessionId },
+          { kind: 'session', sessionId: secondSessionId },
+        ],
+      });
+
+      db = new Database(dbPath);
+      const row = db
+        .prepare('SELECT COUNT(*) AS total FROM imm_sessions WHERE video_id = ?')
+        .get(videoId) as { total: number };
+      assert.equal(row.total, 0);
+    } finally {
+      runtime.destroy();
+      try {
+        db.close();
+      } catch {
+        // The setup connection is already closed before the worker starts.
+      }
+      fs.rmSync(tempDir, { recursive: true, force: true });
+    }
+  },
+);
+
+test('worker runtime warns before falling back when no emitted worker is available', async () => {
+  const warnings: unknown[][] = [];
+  const fallbackTasks: unknown[] = [];
+  const runtime = new DeleteMaintenanceWorkerRuntime({
+    resolveWorkerPath: () => null,
+    warn: (...args) => warnings.push(args),
+    executeFallback: (_dbPath, task) => fallbackTasks.push(task),
+  });
+
+  await runtime.run('/tmp/fallback.sqlite', { kind: 'session', sessionId: 1 });
+
+  assert.equal(warnings.length, 1);
+  assert.match(String(warnings[0]?.[0]), /worker unavailable/i);
+  assert.deepEqual(fallbackTasks, [{ kind: 'session', sessionId: 1 }]);
+});
+
+test('worker runtime terminates a worker after successful settlement', async () => {
+  const { worker, listeners, terminationState } = createFakeWorker();
+  const runtime = new DeleteMaintenanceWorkerRuntime({
+    resolveWorkerPath: () => '/tmp/delete-worker.js',
+    createWorker: async () => worker,
+  });
+
+  const result = runtime.run('/tmp/test.sqlite', { kind: 'session', sessionId: 1 });
+  await new Promise<void>((resolve) => setTimeout(resolve, 0));
+  listeners.get('message')?.({ ok: true } as never);
+  await result;
+
+  assert.equal(terminationState.calls, 1);
+});
+
+test('worker runtime terminates a worker after failed settlement', async () => {
+  const { worker, listeners, terminationState } = createFakeWorker();
+  const runtime = new DeleteMaintenanceWorkerRuntime({
+    resolveWorkerPath: () => '/tmp/delete-worker.js',
+    createWorker: async () => worker,
+  });
+
+  const result = runtime.run('/tmp/test.sqlite', { kind: 'session', sessionId: 1 });
+  await new Promise<void>((resolve) => setTimeout(resolve, 0));
+  listeners.get('error')?.(new Error('worker failed') as never);
+
+  await assert.rejects(result, /worker failed/);
+  assert.equal(terminationState.calls, 1);
+});
+
+test('worker runtime terminates a worker created after shutdown begins', async () => {
+  const { worker, listeners, terminationState } = createFakeWorker();
+  const createGate: { resolve?: (worker: FakeWorker) => void } = {};
+  const fallbackTasks: unknown[] = [];
+  const runtime = new DeleteMaintenanceWorkerRuntime({
+    resolveWorkerPath: () => '/tmp/delete-worker.js',
+    createWorker: () =>
+      new Promise((resolve) => {
+        createGate.resolve = resolve;
+      }),
+    executeFallback: (_dbPath, task) => fallbackTasks.push(task),
+  });
+
+  const result = runtime.run('/tmp/test.sqlite', { kind: 'session', sessionId: 1 });
+  await new Promise<void>((resolve) => setTimeout(resolve, 0));
+  runtime.destroy();
+  createGate.resolve?.(worker);
+
+  await assert.rejects(result, /shut down/);
+  assert.equal(terminationState.calls, 1);
+  assert.equal(listeners.size, 0);
+  assert.deepEqual(fallbackTasks, []);
+});
+
+test('worker runtime does not fall back when worker creation fails during shutdown', async () => {
+  const createGate: { reject?: (error: Error) => void } = {};
+  const fallbackTasks: unknown[] = [];
+  const runtime = new DeleteMaintenanceWorkerRuntime({
+    resolveWorkerPath: () => '/tmp/delete-worker.js',
+    createWorker: () =>
+      new Promise((_resolve, reject) => {
+        createGate.reject = reject;
+      }),
+    executeFallback: (_dbPath, task) => fallbackTasks.push(task),
+  });
+
+  const result = runtime.run('/tmp/test.sqlite', { kind: 'session', sessionId: 1 });
+  await new Promise<void>((resolve) => setTimeout(resolve, 0));
+  runtime.destroy();
+  createGate.reject?.(new Error('creation failed'));
+
+  await assert.rejects(result, /shut down/);
+  assert.deepEqual(fallbackTasks, []);
+});

src/core/services/immersion-tracker/delete-maintenance-worker-runtime.ts

@@ -0,0 +1,121 @@
+import fs from 'node:fs';
+import path from 'node:path';
+import { createLogger } from '../../../logger';
+import { executeDeleteMaintenanceTask, type DeleteMaintenanceTask } from './delete-maintenance';
+
+interface DeleteMaintenanceWorkerResponse {
+  ok?: unknown;
+  error?: unknown;
+}
+
+export type RunDeleteMaintenanceTask = (
+  dbPath: string,
+  task: DeleteMaintenanceTask,
+) => Promise<void>;
+
+interface DeleteMaintenanceWorkerHandle {
+  once(event: 'message', listener: (message: DeleteMaintenanceWorkerResponse) => void): this;
+  once(event: 'error', listener: (error: Error) => void): this;
+  once(event: 'exit', listener: (code: number) => void): this;
+  terminate(): Promise<number>;
+}
+
+interface DeleteMaintenanceWorkerRuntimeOptions {
+  resolveWorkerPath?: () => string | null;
+  createWorker?: (
+    workerPath: string,
+    workerData: { dbPath: string; task: DeleteMaintenanceTask },
+  ) => Promise<DeleteMaintenanceWorkerHandle>;
+  executeFallback?: typeof executeDeleteMaintenanceTask;
+  warn?: (message: string, ...meta: unknown[]) => void;
+}
+
+export function resolveDeleteMaintenanceWorkerPath(): string | null {
+  const workerPath = path.join(__dirname, 'delete-maintenance-worker-thread.js');
+  return fs.existsSync(workerPath) ? workerPath : null;
+}
+
+const logger = createLogger('main:immersion-tracker:delete-worker');
+
+export class DeleteMaintenanceWorkerRuntime {
+  private readonly activeWorkers = new Set<DeleteMaintenanceWorkerHandle>();
+  private destroyed = false;
+
+  constructor(private readonly options: DeleteMaintenanceWorkerRuntimeOptions = {}) {}
+
+  async run(dbPath: string, task: DeleteMaintenanceTask): Promise<void> {
+    if (this.destroyed) {
+      throw new Error('Delete maintenance worker is shut down');
+    }
+
+    let worker: DeleteMaintenanceWorkerHandle;
+    try {
+      const workerPath = (this.options.resolveWorkerPath ?? resolveDeleteMaintenanceWorkerPath)();
+      if (!workerPath) throw new Error('Emitted delete-maintenance worker module was not found');
+      const createWorker =
+        this.options.createWorker ??
+        (async (resolvedPath, workerData) => {
+          const { Worker } = await import('node:worker_threads');
+          return new Worker(resolvedPath, { workerData });
+        });
+      worker = await createWorker(workerPath, { dbPath, task });
+    } catch (error) {
+      if (this.destroyed) {
+        throw new Error('Delete maintenance worker is shut down');
+      }
+      (this.options.warn ?? logger.warn)(
+        'Delete maintenance worker unavailable; running maintenance on the current thread',
+        error,
+      );
+      (this.options.executeFallback ?? executeDeleteMaintenanceTask)(dbPath, task);
+      return;
+    }
+
+    if (this.destroyed) {
+      await worker.terminate().catch(() => undefined);
+      throw new Error('Delete maintenance worker is shut down');
+    }
+
+    await new Promise<void>((resolve, reject) => {
+      let settled = false;
+      this.activeWorkers.add(worker);
+
+      const settle = (error?: Error) => {
+        if (settled) return;
+        settled = true;
+        this.activeWorkers.delete(worker);
+        if (error) reject(error);
+        else resolve();
+        void worker.terminate();
+      };
+
+      worker.once('message', (message: DeleteMaintenanceWorkerResponse) => {
+        if (message.ok === true) {
+          settle();
+          return;
+        }
+        const detail = typeof message.error === 'string' ? message.error : 'unknown worker error';
+        settle(new Error(`Delete maintenance failed: ${detail}`));
+      });
+      worker.once('error', (error) => settle(error));
+      worker.once('exit', (code) => {
+        settle(
+          new Error(
+            code === 0
+              ? 'Delete maintenance worker exited without a response'
+              : `Delete maintenance worker exited with code ${code}`,
+          ),
+        );
+      });
+    });
+  }
+
+  destroy(): void {
+    if (this.destroyed) return;
+    this.destroyed = true;
+    for (const worker of this.activeWorkers) {
+      void worker.terminate();
+    }
+    this.activeWorkers.clear();
+  }
+}

src/core/services/immersion-tracker/delete-maintenance-worker-thread.ts

@@ -0,0 +1,22 @@
+import { parentPort, workerData } from 'node:worker_threads';
+import { executeDeleteMaintenanceTask, type DeleteMaintenanceTask } from './delete-maintenance';
+
+interface DeleteMaintenanceWorkerData {
+  dbPath: string;
+  task: DeleteMaintenanceTask;
+}
+
+if (!parentPort) {
+  throw new Error('delete maintenance worker missing parent port');
+}
+
+const port = parentPort;
+const request = workerData as DeleteMaintenanceWorkerData;
+
+try {
+  executeDeleteMaintenanceTask(request.dbPath, request.task);
+  port.postMessage({ ok: true });
+} catch (error) {
+  const message = error instanceof Error ? error.message : String(error);
+  port.postMessage({ ok: false, error: message });
+}

src/core/services/immersion-tracker/delete-maintenance.ts

@@ -0,0 +1,47 @@
+import { Database } from './sqlite';
+import { applyPragmas } from './storage';
+import { deleteAnime, deleteSession, deleteSessions, deleteVideo } from './query-maintenance';
+import {
+  deleteMaintenanceBatch,
+  type DeleteMaintenanceOperation,
+} from './query-delete-maintenance';
+
+export type { DeleteMaintenanceOperation } from './query-delete-maintenance';
+
+export type DeleteMaintenanceTask =
+  | DeleteMaintenanceOperation
+  | { kind: 'batch'; tasks: DeleteMaintenanceOperation[] };
+
+function executeDeleteMaintenanceOperation(
+  db: InstanceType<typeof Database>,
+  task: DeleteMaintenanceOperation,
+): void {
+  switch (task.kind) {
+    case 'session':
+      deleteSession(db, task.sessionId);
+      return;
+    case 'sessions':
+      deleteSessions(db, task.sessionIds);
+      return;
+    case 'video':
+      deleteVideo(db, task.videoId);
+      return;
+    case 'anime':
+      deleteAnime(db, task.animeId);
+      return;
+  }
+}
+
+export function executeDeleteMaintenanceTask(dbPath: string, task: DeleteMaintenanceTask): void {
+  const db = new Database(dbPath);
+  try {
+    applyPragmas(db);
+    if (task.kind === 'batch') {
+      deleteMaintenanceBatch(db, task.tasks);
+      return;
+    }
+    executeDeleteMaintenanceOperation(db, task);
+  } finally {
+    db.close();
+  }
+}

src/core/services/immersion-tracker/jellyfin-link-repair.ts

@@ -8,6 +8,8 @@ import type { JellyfinLinkRepairSummary } from './types';
 type LegacyJellyfinVideoRow = {
   video_id: number;
   video_key: string;
+  anime_id: number | null;
+  anime_assignment_locked: number;
   source_url: string | null;
   canonical_title: string;
 };
@@ -15,6 +17,7 @@ type LegacyJellyfinVideoRow = {
 type JellyfinTargetVideoRow = {
   video_id: number;
   anime_id: number | null;
+  anime_assignment_locked: number;
   canonical_title: string;
   parsed_basename: string | null;
   parsed_title: string | null;
@@ -258,7 +261,13 @@ export function repairJellyfinStreamVideoLinks(db: DatabaseSync): JellyfinLinkRe
   const candidates = db
     .prepare(
       `
-        SELECT video_id, video_key, source_url, canonical_title
+        SELECT
+          video_id,
+          video_key,
+          anime_id,
+          anime_assignment_locked,
+          source_url,
+          canonical_title
         FROM imm_videos
         WHERE source_type = 2
           AND (
@@ -310,6 +319,7 @@ export function repairJellyfinStreamVideoLinks(db: DatabaseSync): JellyfinLinkRe
             SELECT
               video_id,
               anime_id,
+              anime_assignment_locked,
               canonical_title,
               parsed_basename,
               parsed_title,
@@ -357,12 +367,17 @@ export function repairJellyfinStreamVideoLinks(db: DatabaseSync): JellyfinLinkRe
         continue;
       }
 
+      const assignmentAnimeId =
+        candidate.anime_assignment_locked === 1 ? candidate.anime_id : target.anime_id;
+      const assignmentLocked =
+        candidate.anime_assignment_locked === 1 || target.anime_assignment_locked === 1 ? 1 : 0;
       db.prepare(
         `
           UPDATE imm_videos
           SET
             video_key = ?,
             anime_id = ?,
+            anime_assignment_locked = ?,
             canonical_title = ?,
             source_url = ?,
             parsed_basename = ?,
@@ -377,7 +392,8 @@ export function repairJellyfinStreamVideoLinks(db: DatabaseSync): JellyfinLinkRe
         `,
       ).run(
         sanitizedVideoKey,
-        target.anime_id,
+        assignmentAnimeId,
+        assignmentLocked,
         target.canonical_title,
         statsUrl,
         target.parsed_basename,
@@ -390,14 +406,14 @@ export function repairJellyfinStreamVideoLinks(db: DatabaseSync): JellyfinLinkRe
         currentTimestamp,
         candidate.video_id,
       );
-      if (target.anime_id !== null) {
+      if (assignmentAnimeId !== null) {
         db.prepare(
           `
             UPDATE imm_subtitle_lines
             SET anime_id = ?, LAST_UPDATE_DATE = ?
             WHERE video_id = ?
           `,
-        ).run(target.anime_id, currentTimestamp, candidate.video_id);
+        ).run(assignmentAnimeId, currentTimestamp, candidate.video_id);
       }
       summary.repaired += 1;
     }

src/core/services/immersion-tracker/lifetime.ts

@@ -708,6 +708,87 @@ export function rebuildLifetimeSummariesInTransaction(
   return rebuildLifetimeSummariesInternal(db, rebuiltAtMs);
 }
 
+/**
+ * Re-derive every per-anime lifetime row from the per-video summaries after
+ * episodes changed owners (merge, move, season repair).
+ *
+ * Deliberately NOT a full rebuild: {@link rebuildLifetimeSummariesInTransaction}
+ * recomputes from raw sessions, which are pruned after the retention window, so
+ * it silently truncates lifetime history. `imm_lifetime_media` is keyed by
+ * video and survives repointing, so aggregating it preserves all-time totals;
+ * `imm_lifetime_global` only needs `anime_completed` refreshed because moving
+ * attribution between entries cannot change the global counters.
+ *
+ * Assumes the caller holds a write transaction; use
+ * {@link recomputeLifetimeAnimeAggregates} otherwise.
+ */
+export function recomputeLifetimeAnimeAggregatesInTransaction(db: DatabaseSync): void {
+  const updatedAt = toDbTimestamp(nowMs());
+  db.exec('DELETE FROM imm_lifetime_anime');
+  db.prepare(
+    `
+    INSERT INTO imm_lifetime_anime (
+      anime_id,
+      total_sessions,
+      total_active_ms,
+      total_cards,
+      total_lines_seen,
+      total_tokens_seen,
+      episodes_started,
+      episodes_completed,
+      first_watched_ms,
+      last_watched_ms,
+      CREATED_DATE,
+      LAST_UPDATE_DATE
+    )
+    SELECT
+      v.anime_id,
+      COALESCE(SUM(m.total_sessions), 0),
+      COALESCE(SUM(m.total_active_ms), 0),
+      COALESCE(SUM(m.total_cards), 0),
+      COALESCE(SUM(m.total_lines_seen), 0),
+      COALESCE(SUM(m.total_tokens_seen), 0),
+      COUNT(*),
+      COUNT(CASE WHEN m.completed > 0 THEN 1 END),
+      MIN(m.first_watched_ms),
+      MAX(m.last_watched_ms),
+      ?,
+      ?
+    FROM imm_lifetime_media m
+    JOIN imm_videos v ON v.video_id = m.video_id
+    WHERE v.anime_id IS NOT NULL
+    GROUP BY v.anime_id
+    `,
+  ).run(updatedAt, updatedAt);
+  db.prepare(
+    `
+    UPDATE imm_lifetime_global
+    SET
+      anime_completed = (
+        SELECT COUNT(*)
+        FROM imm_lifetime_anime la
+        JOIN imm_anime a ON a.anime_id = la.anime_id
+        WHERE a.episodes_total IS NOT NULL
+          AND a.episodes_total > 0
+          AND la.episodes_completed >= a.episodes_total
+      ),
+      LAST_UPDATE_DATE = ?
+    WHERE global_id = 1
+    `,
+  ).run(updatedAt);
+}
+
+export function recomputeLifetimeAnimeAggregates(db: DatabaseSync): void {
+  db.exec('BEGIN IMMEDIATE');
+  try {
+    recomputeLifetimeAnimeAggregatesInTransaction(db);
+    db.exec('COMMIT');
+  } catch (error) {
+    db.exec('ROLLBACK');
+    throw error;
+  }
+}
+
 export function reconcileStaleActiveSessions(db: DatabaseSync): number {
   const sessions = getRetainedStaleActiveSessions(db);
   if (sessions.length === 0) {

src/core/services/immersion-tracker/query-delete-maintenance.ts

@@ -0,0 +1,196 @@
+import type { DatabaseSync } from './sqlite';
+import { rebuildLifetimeSummariesInTransaction } from './lifetime';
+import { getRollupGroupsForSessions, refreshRollupsForGroupsInTransaction } from './maintenance';
+import {
+  applyLexicalRemovals,
+  cleanupUnusedCoverArtBlobHash,
+  deleteSessionsByIds,
+  forEachIdChunk,
+  makePlaceholders,
+  planLexicalRemovalsForSessions,
+  SQLITE_ID_CHUNK_SIZE,
+  type LexicalRemovalPlan,
+} from './query-shared';
+
+export type DeleteMaintenanceOperation =
+  | { kind: 'session'; sessionId: number }
+  | { kind: 'sessions'; sessionIds: number[] }
+  | { kind: 'video'; videoId: number }
+  | { kind: 'anime'; animeId: number };
+
+function addOperationTargets(
+  operations: DeleteMaintenanceOperation[],
+  sessionIds: Set<number>,
+  videoIds: Set<number>,
+  animeIds: Set<number>,
+): void {
+  for (const operation of operations) {
+    switch (operation.kind) {
+      case 'session':
+        sessionIds.add(operation.sessionId);
+        break;
+      case 'sessions':
+        for (const sessionId of operation.sessionIds) sessionIds.add(sessionId);
+        break;
+      case 'video':
+        videoIds.add(operation.videoId);
+        break;
+      case 'anime':
+        animeIds.add(operation.animeId);
+        break;
+    }
+  }
+}
+
+function selectIds(
+  db: DatabaseSync,
+  buildSql: (placeholders: string) => string,
+  params: number[],
+  column: string,
+): number[] {
+  if (params.length === 0) return [];
+  const ids: number[] = [];
+  forEachIdChunk(params, (chunk) => {
+    const rows = db.prepare(buildSql(makePlaceholders(chunk))).all(...chunk) as Array<
+      Record<string, number>
+    >;
+    for (const row of rows) ids.push(row[column]!);
+  });
+  return ids;
+}
+
+function planLexicalRemovalsInChunks(db: DatabaseSync, sessionIds: number[]): LexicalRemovalPlan {
+  const combined: LexicalRemovalPlan = { words: [], kanji: [] };
+  const merge = (target: LexicalRemovalPlan['words'], source: LexicalRemovalPlan['words']) => {
+    const byId = new Map(target.map((entry) => [entry.id, entry]));
+    for (const entry of source) {
+      const existing = byId.get(entry.id);
+      if (!existing) {
+        const added = { ...entry };
+        target.push(added);
+        byId.set(entry.id, added);
+        continue;
+      }
+      existing.removedFrequency += entry.removedFrequency;
+      if (
+        entry.removedFirstSeenMs !== null &&
+        (existing.removedFirstSeenMs === null ||
+          entry.removedFirstSeenMs < existing.removedFirstSeenMs)
+      ) {
+        existing.removedFirstSeenMs = entry.removedFirstSeenMs;
+      }
+      if (
+        entry.removedLastSeenMs !== null &&
+        (existing.removedLastSeenMs === null ||
+          entry.removedLastSeenMs > existing.removedLastSeenMs)
+      ) {
+        existing.removedLastSeenMs = entry.removedLastSeenMs;
+      }
+    }
+  };
+
+  forEachIdChunk(sessionIds, (chunk) => {
+    const plan = planLexicalRemovalsForSessions(db, chunk);
+    merge(combined.words, plan.words);
+    merge(combined.kanji, plan.kanji);
+  });
+  return combined;
+}
+
+export function deleteMaintenanceBatch(
+  db: DatabaseSync,
+  operations: DeleteMaintenanceOperation[],
+): void {
+  if (operations.length === 0) return;
+
+  db.exec('BEGIN IMMEDIATE');
+  try {
+    const sessionIds = new Set<number>();
+    const videoIds = new Set<number>();
+    const animeIds = new Set<number>();
+    addOperationTargets(operations, sessionIds, videoIds, animeIds);
+
+    const animeIdList = [...animeIds];
+    for (const videoId of selectIds(
+      db,
+      (placeholders) => `SELECT video_id FROM imm_videos WHERE anime_id IN (${placeholders})`,
+      animeIdList,
+      'video_id',
+    )) {
+      videoIds.add(videoId);
+    }
+
+    const videoIdList = [...videoIds];
+    for (const sessionId of selectIds(
+      db,
+      (placeholders) => `SELECT session_id FROM imm_sessions WHERE video_id IN (${placeholders})`,
+      videoIdList,
+      'session_id',
+    )) {
+      sessionIds.add(sessionId);
+    }
+
+    const sessionIdList = [...sessionIds];
+    const lexicalRemovals = planLexicalRemovalsInChunks(db, sessionIdList);
+    const affectedRollupGroups = sessionIdList
+      .flatMap((_, index) =>
+        index % SQLITE_ID_CHUNK_SIZE === 0
+          ? getRollupGroupsForSessions(db, sessionIdList.slice(index, index + SQLITE_ID_CHUNK_SIZE))
+          : [],
+      )
+      .filter((group) => !videoIds.has(group.videoId));
+    const coverBlobHashes = new Set<string>();
+    if (videoIdList.length > 0) {
+      forEachIdChunk(videoIdList, (chunk) => {
+        const placeholders = makePlaceholders(chunk);
+        const artRows = db
+          .prepare(
+            `SELECT cover_blob_hash AS coverBlobHash
+             FROM imm_media_art
+             WHERE video_id IN (${placeholders}) AND cover_blob_hash IS NOT NULL`,
+          )
+          .all(...chunk) as Array<{ coverBlobHash: string }>;
+        for (const row of artRows) coverBlobHashes.add(row.coverBlobHash);
+      });
+
+      deleteSessionsByIds(db, sessionIdList);
+      forEachIdChunk(videoIdList, (chunk) => {
+        const placeholders = makePlaceholders(chunk);
+        db.prepare(`DELETE FROM imm_subtitle_lines WHERE video_id IN (${placeholders})`).run(
+          ...chunk,
+        );
+        db.prepare(`DELETE FROM imm_daily_rollups WHERE video_id IN (${placeholders})`).run(
+          ...chunk,
+        );
+        db.prepare(`DELETE FROM imm_monthly_rollups WHERE video_id IN (${placeholders})`).run(
+          ...chunk,
+        );
+        db.prepare(`DELETE FROM imm_media_art WHERE video_id IN (${placeholders})`).run(...chunk);
+        db.prepare(`DELETE FROM imm_videos WHERE video_id IN (${placeholders})`).run(...chunk);
+      });
+    } else {
+      deleteSessionsByIds(db, sessionIdList);
+    }
+
+    for (const coverBlobHash of coverBlobHashes) {
+      cleanupUnusedCoverArtBlobHash(db, coverBlobHash);
+    }
+    if (animeIdList.length > 0) {
+      forEachIdChunk(animeIdList, (chunk) => {
+        const placeholders = makePlaceholders(chunk);
+        db.prepare(`DELETE FROM imm_lifetime_anime WHERE anime_id IN (${placeholders})`).run(
+          ...chunk,
+        );
+        db.prepare(`DELETE FROM imm_anime WHERE anime_id IN (${placeholders})`).run(...chunk);
+      });
+    }
+
+    applyLexicalRemovals(db, lexicalRemovals);
+    rebuildLifetimeSummariesInTransaction(db);
+    refreshRollupsForGroupsInTransaction(db, affectedRollupGroups);
+    db.exec('COMMIT');
+  } catch (error) {
+    db.exec('ROLLBACK');
+    throw error;
+  }
+}

src/core/services/immersion-tracker/query-maintenance.ts

@@ -1,7 +1,10 @@
 import { createHash } from 'node:crypto';
 import type { DatabaseSync } from './sqlite';
 import { buildCoverBlobReference, normalizeCoverBlobBytes } from './storage';
-import { rebuildLifetimeSummaries, rebuildLifetimeSummariesInTransaction } from './lifetime';
+import {
+  recomputeLifetimeAnimeAggregates,
+  rebuildLifetimeSummariesInTransaction,
+} from './lifetime';
 import { getRollupGroupsForSessions, refreshRollupsForGroupsInTransaction } from './maintenance';
 import { nowMs } from './time';
 import { resolveAnimeAnilistConflict } from './anime-season-repair';
@@ -418,6 +421,7 @@ export function updateAnimeAnilistInfo(
     titleEnglish: string | null;
     titleNative: string | null;
     episodesTotal: number | null;
+    exactTitleMatch?: boolean;
   },
 ): void {
   const row = db.prepare('SELECT anime_id FROM imm_videos WHERE video_id = ?').get(videoId) as {
@@ -425,7 +429,11 @@ export function updateAnimeAnilistInfo(
   } | null;
   if (!row?.anime_id) return;
 
-  const repair = resolveAnimeAnilistConflict(db, row.anime_id, info.anilistId);
+  const repair = resolveAnimeAnilistConflict(db, row.anime_id, info.anilistId, {
+    matchConfidence:
+      info.exactTitleMatch === true ? 'exact' : info.exactTitleMatch === false ? 'weak' : undefined,
+  });
+  if (repair.mergeRecommended || repair.anilistAssignmentBlocked) return;
   const targetRow = db
     .prepare('SELECT anime_id FROM imm_videos WHERE video_id = ?')
     .get(videoId) as {
@@ -455,7 +463,7 @@ export function updateAnimeAnilistInfo(
     targetRow.anime_id,
   );
   if (repair.movedVideos > 0 || repair.deletedAnimeRows > 0) {
-    rebuildLifetimeSummaries(db);
+    recomputeLifetimeAnimeAggregates(db);
   }
 }
 

src/core/services/immersion-tracker/query-shared.ts

@@ -80,6 +80,14 @@ export function makePlaceholders(values: number[]): string {
   return values.map(() => '?').join(',');
 }
 
+export const SQLITE_ID_CHUNK_SIZE = 1_000;
+
+export function forEachIdChunk(ids: number[], callback: (chunk: number[]) => void): void {
+  for (let start = 0; start < ids.length; start += SQLITE_ID_CHUNK_SIZE) {
+    callback(ids.slice(start, start + SQLITE_ID_CHUNK_SIZE));
+  }
+}
+
 export function resolvedCoverBlobExpr(mediaAlias: string, blobStoreAlias: string): string {
   return `COALESCE(${blobStoreAlias}.cover_blob, CASE WHEN ${mediaAlias}.cover_blob_hash IS NULL THEN ${mediaAlias}.cover_blob ELSE NULL END)`;
 }
@@ -490,17 +498,19 @@ export function deleteSessionsByIds(db: DatabaseSync, sessionIds: number[]): voi
     return;
   }
 
-  const placeholders = makePlaceholders(sessionIds);
+  forEachIdChunk(sessionIds, (chunk) => {
-  db.prepare(`DELETE FROM imm_subtitle_lines WHERE session_id IN (${placeholders})`).run(
+    const placeholders = makePlaceholders(chunk);
-    ...sessionIds,
+    db.prepare(`DELETE FROM imm_subtitle_lines WHERE session_id IN (${placeholders})`).run(
-  );
+      ...chunk,
-  db.prepare(`DELETE FROM imm_session_telemetry WHERE session_id IN (${placeholders})`).run(
+    );
-    ...sessionIds,
+    db.prepare(`DELETE FROM imm_session_telemetry WHERE session_id IN (${placeholders})`).run(
-  );
+      ...chunk,
-  db.prepare(`DELETE FROM imm_session_events WHERE session_id IN (${placeholders})`).run(
+    );
-    ...sessionIds,
+    db.prepare(`DELETE FROM imm_session_events WHERE session_id IN (${placeholders})`).run(
-  );
+      ...chunk,
-  db.prepare(`DELETE FROM imm_sessions WHERE session_id IN (${placeholders})`).run(...sessionIds);
+    );
+    db.prepare(`DELETE FROM imm_sessions WHERE session_id IN (${placeholders})`).run(...chunk);
+  });
 }
 
 export function toDbMs(ms: number | bigint): bigint {

src/core/services/immersion-tracker/storage-session.test.ts

@@ -20,6 +20,7 @@ import {
 } from './storage';
 import {
   EVENT_SUBTITLE_LINE,
+  SCHEMA_VERSION,
   SESSION_STATUS_ENDED,
   SOURCE_TYPE_LOCAL,
   SOURCE_TYPE_REMOTE,
@@ -132,6 +133,7 @@ test('ensureSchema creates immersion core tables', () => {
     assert.ok(videoColumns.has('parser_source'));
     assert.ok(videoColumns.has('parser_confidence'));
     assert.ok(videoColumns.has('parse_metadata_json'));
+    assert.ok(videoColumns.has('anime_assignment_locked'));
 
     const mediaArtColumns = new Set(
       (
@@ -155,6 +157,33 @@ test('ensureSchema creates immersion core tables', () => {
   }
 });
 
+test('ensureSchema adds manual assignment locks when upgrading the previous schema', () => {
+  const dbPath = makeDbPath();
+  const db = new Database(dbPath);
+
+  try {
+    ensureSchema(db);
+    db.exec('ALTER TABLE imm_videos DROP COLUMN anime_assignment_locked');
+    db.prepare('UPDATE imm_schema_version SET schema_version = ?').run(SCHEMA_VERSION - 1);
+
+    ensureSchema(db);
+
+    const columns = new Set(
+      (db.prepare('PRAGMA table_info(imm_videos)').all() as Array<{ name: string }>).map(
+        (row) => row.name,
+      ),
+    );
+    assert.ok(columns.has('anime_assignment_locked'));
+    const version = db
+      .prepare('SELECT MAX(schema_version) AS version FROM imm_schema_version')
+      .get() as { version: number };
+    assert.equal(version.version, SCHEMA_VERSION);
+  } finally {
+    db.close();
+    cleanupDbPath(dbPath);
+  }
+});
+
 test('stats excluded words are replaced and read from sqlite storage', () => {
   const dbPath = makeDbPath();
   const db = new Database(dbPath);
@@ -807,6 +836,7 @@ test('ensureSchema migrates legacy videos and backfills anime metadata from file
     assert.ok(videoColumns.has('parser_source'));
     assert.ok(videoColumns.has('parser_confidence'));
     assert.ok(videoColumns.has('parse_metadata_json'));
+    assert.ok(videoColumns.has('anime_assignment_locked'));
 
     const animeRows = db
       .prepare('SELECT canonical_title FROM imm_anime ORDER BY canonical_title')

src/core/services/immersion-tracker/storage.ts

@@ -1,5 +1,7 @@
 import { createHash } from 'node:crypto';
+import path from 'node:path';
 import { parseMediaInfo } from '../../../jimaku/utils';
+import { normalizeTitleIdentity } from '../../utils/title-normalization';
 import type { DatabaseSync } from './sqlite';
 import { nowMs } from './time';
 import { SCHEMA_VERSION } from './types';
@@ -319,14 +321,7 @@ export function applyPragmas(db: DatabaseSync): void {
   db.exec(`PRAGMA journal_size_limit = ${WAL_JOURNAL_SIZE_LIMIT_BYTES}`);
 }
 
-export function normalizeAnimeIdentityKey(title: string): string {
+export const normalizeAnimeIdentityKey = normalizeTitleIdentity;
-  return title
-    .normalize('NFKC')
-    .toLowerCase()
-    .replace(/[^\p{L}\p{N}]+/gu, ' ')
-    .trim()
-    .replace(/\s+/g, ' ');
-}
 
 function normalizeSeasonScope(value: number | null | undefined): number | null {
   if (typeof value !== 'number' || !Number.isSafeInteger(value) || value <= 0) {
@@ -530,6 +525,36 @@ function ensureStatsExcludedWordsTable(db: DatabaseSync): void {
   `);
 }
 
+function ensureAnimeMergeTables(db: DatabaseSync): void {
+  db.exec(`
+    CREATE TABLE IF NOT EXISTS imm_anime_title_aliases(
+      normalized_title_key TEXT PRIMARY KEY,
+      anime_id INTEGER NOT NULL,
+      CREATED_DATE TEXT,
+      LAST_UPDATE_DATE TEXT,
+      FOREIGN KEY(anime_id) REFERENCES imm_anime(anime_id) ON DELETE CASCADE
+    );
+    CREATE INDEX IF NOT EXISTS idx_anime_title_aliases_anime_id
+    ON imm_anime_title_aliases(anime_id);
+
+    CREATE TABLE IF NOT EXISTS imm_anime_merge_recommendations(
+      recommendation_id INTEGER PRIMARY KEY AUTOINCREMENT,
+      first_anime_id INTEGER NOT NULL,
+      second_anime_id INTEGER NOT NULL,
+      anilist_id INTEGER NOT NULL,
+      status TEXT NOT NULL DEFAULT 'pending' CHECK(status IN ('pending', 'dismissed')),
+      CREATED_DATE TEXT,
+      LAST_UPDATE_DATE TEXT,
+      CHECK(first_anime_id < second_anime_id),
+      UNIQUE(first_anime_id, second_anime_id, anilist_id),
+      FOREIGN KEY(first_anime_id) REFERENCES imm_anime(anime_id) ON DELETE CASCADE,
+      FOREIGN KEY(second_anime_id) REFERENCES imm_anime(anime_id) ON DELETE CASCADE
+    );
+    CREATE INDEX IF NOT EXISTS idx_anime_merge_recommendations_status
+    ON imm_anime_merge_recommendations(status, recommendation_id);
+  `);
+}
+
 export function getOrCreateAnimeRecord(db: DatabaseSync, input: AnimeRecordInput): number {
   const seasonScope = normalizeSeasonScope(input.seasonScope);
   const identityTitle = buildSeasonScopedAnimeTitle(input.parsedTitle, seasonScope);
@@ -550,8 +575,14 @@ export function getOrCreateAnimeRecord(db: DatabaseSync, input: AnimeRecordInput
   const byNormalizedTitle = db
     .prepare('SELECT anime_id FROM imm_anime WHERE normalized_title_key = ?')
     .get(normalizedTitleKey) as { anime_id: number } | null;
-  const existing = byAnilistId ?? byNormalizedTitle;
+  const byTitleAlias = db
+    .prepare('SELECT anime_id FROM imm_anime_title_aliases WHERE normalized_title_key = ?')
+    .get(normalizedTitleKey) as { anime_id: number } | null;
+  const existing = byAnilistId ?? byNormalizedTitle ?? byTitleAlias;
   if (existing?.anime_id) {
+    // An alias remembers an intentionally merged-away spelling. Reusing it
+    // must not rename the survivor back to that discarded display title.
+    const canonicalTitleUpdate = byAnilistId || byNormalizedTitle ? canonicalTitle : null;
     db.prepare(
       `
         UPDATE imm_anime
@@ -566,7 +597,7 @@ export function getOrCreateAnimeRecord(db: DatabaseSync, input: AnimeRecordInput
         WHERE anime_id = ?
       `,
     ).run(
-      canonicalTitle,
+      canonicalTitleUpdate,
       input.anilistId,
       input.titleRomaji,
       input.titleEnglish,
@@ -618,7 +649,10 @@ export function linkVideoToAnimeRecord(
     `
       UPDATE imm_videos
       SET
-        anime_id = ?,
+        anime_id = CASE
+          WHEN anime_assignment_locked = 1 THEN anime_id
+          ELSE ?
+        END,
         parsed_basename = ?,
         parsed_title = ?,
         parsed_season = ?,
@@ -643,6 +677,67 @@ export function linkVideoToAnimeRecord(
   );
 }
 
+export function getManualAnimeAssignment(db: DatabaseSync, videoId: number): number | null {
+  const row = db
+    .prepare(
+      `
+        SELECT anime_id AS animeId
+        FROM imm_videos
+        WHERE video_id = ?
+          AND anime_assignment_locked = 1
+      `,
+    )
+    .get(videoId) as { animeId: number | null } | null;
+  return row?.animeId ?? null;
+}
+
+/**
+ * A manual correction in the same folder is a useful grouping hint, but only
+ * when every season-compatible correction agrees on the destination.
+ */
+export function findManualDirectoryAnimeAssignment(
+  db: DatabaseSync,
+  videoId: number,
+  mediaPath: string,
+  parsedSeason: number | null,
+): number | null {
+  const directory = path.dirname(path.resolve(mediaPath));
+  const rows = db
+    .prepare(
+      `
+        SELECT
+          anime_id AS animeId,
+          source_path AS sourcePath,
+          parsed_season AS parsedSeason
+        FROM imm_videos
+        WHERE video_id != ?
+          AND anime_assignment_locked = 1
+          AND anime_id IS NOT NULL
+          AND source_path IS NOT NULL
+      `,
+    )
+    .all(videoId) as Array<{
+    animeId: number;
+    sourcePath: string;
+    parsedSeason: number | null;
+  }>;
+
+  const candidates = new Set<number>();
+  for (const row of rows) {
+    if (path.dirname(path.resolve(row.sourcePath)) !== directory) {
+      continue;
+    }
+    if (parsedSeason !== null && row.parsedSeason !== null && parsedSeason !== row.parsedSeason) {
+      continue;
+    }
+    candidates.add(row.animeId);
+    if (candidates.size > 1) {
+      return null;
+    }
+  }
+  return candidates.values().next().value ?? null;
+}
+
 export function linkYoutubeVideoToAnimeRecord(
   db: DatabaseSync,
   videoId: number,
@@ -751,6 +846,7 @@ export function ensureSchema(db: DatabaseSync): void {
   if (currentVersion?.schema_version === SCHEMA_VERSION) {
     ensureLifetimeSummaryTables(db);
     ensureStatsExcludedWordsTable(db);
+    ensureAnimeMergeTables(db);
     return;
   }
 
@@ -786,6 +882,7 @@ export function ensureSchema(db: DatabaseSync): void {
       parser_source TEXT,
       parser_confidence REAL,
       parse_metadata_json TEXT,
+      anime_assignment_locked INTEGER NOT NULL DEFAULT 0 CHECK(anime_assignment_locked IN (0, 1)),
       watched INTEGER NOT NULL DEFAULT 0,
       duration_ms INTEGER NOT NULL CHECK(duration_ms>=0),
       file_size_bytes INTEGER CHECK(file_size_bytes>=0),
@@ -799,6 +896,13 @@ export function ensureSchema(db: DatabaseSync): void {
       FOREIGN KEY(anime_id) REFERENCES imm_anime(anime_id) ON DELETE SET NULL
     );
   `);
+  addColumnIfMissing(
+    db,
+    'imm_videos',
+    'anime_assignment_locked',
+    'INTEGER NOT NULL DEFAULT 0 CHECK(anime_assignment_locked IN (0, 1))',
+  );
+  ensureAnimeMergeTables(db);
   db.exec(`
     CREATE TABLE IF NOT EXISTS imm_sessions(
       session_id INTEGER PRIMARY KEY AUTOINCREMENT,

src/core/services/immersion-tracker/types.ts

@@ -1,4 +1,4 @@
-export const SCHEMA_VERSION = 19;
+export const SCHEMA_VERSION = 21;
 export const DEFAULT_QUEUE_CAP = 1_000;
 export const DEFAULT_BATCH_SIZE = 25;
 export const DEFAULT_FLUSH_INTERVAL_MS = 500;

src/core/services/index.ts

@@ -1,5 +1,9 @@
 export { Texthooker } from './texthooker';
-export { hasMpvWebsocketPlugin, SubtitleWebSocket } from './subtitle-ws';
+export {
+  hasMpvWebsocketPlugin,
+  isSubtitleAnnotationUpgrade,
+  SubtitleWebSocket,
+} from './subtitle-ws';
 export { registerGlobalShortcuts } from './shortcut';
 export { createIpcDepsRuntime, registerIpcHandlers } from './ipc';
 export { shortcutMatchesInputForLocalFallback } from './shortcut-fallback';

src/core/services/ipc.ts

@@ -1,6 +1,7 @@
 import electron from 'electron';
 import type { BrowserWindow as ElectronBrowserWindow, IpcMainEvent } from 'electron';
 import type {
+  ChangelogSnapshot,
   CompiledSessionBinding,
   ControllerConfigUpdate,
   PlaylistBrowserMutationResult,
@@ -122,6 +123,7 @@ export interface IpcServiceDeps {
   removeCharacterDictionaryManagedEntry?: (mediaId: number) => Promise<unknown>;
   moveCharacterDictionaryManagedEntry?: (mediaId: number, direction: 1 | -1) => Promise<unknown>;
   appendClipboardVideoToQueue: () => { ok: boolean; message: string };
+  getChangelogSnapshot?: (options?: { refresh?: boolean }) => Promise<ChangelogSnapshot>;
   getPlaylistBrowserSnapshot: () => Promise<PlaylistBrowserSnapshot>;
   appendPlaylistBrowserFile: (filePath: string) => Promise<PlaylistBrowserMutationResult>;
   playPlaylistBrowserIndex: (index: number) => Promise<PlaylistBrowserMutationResult>;
@@ -297,6 +299,7 @@ export interface IpcDepsRuntimeOptions {
   removeCharacterDictionaryManagedEntry?: (mediaId: number) => Promise<unknown>;
   moveCharacterDictionaryManagedEntry?: (mediaId: number, direction: 1 | -1) => Promise<unknown>;
   appendClipboardVideoToQueue: () => { ok: boolean; message: string };
+  getChangelogSnapshot?: (options?: { refresh?: boolean }) => Promise<ChangelogSnapshot>;
   getPlaylistBrowserSnapshot: () => Promise<PlaylistBrowserSnapshot>;
   appendPlaylistBrowserFile: (filePath: string) => Promise<PlaylistBrowserMutationResult>;
   playPlaylistBrowserIndex: (index: number) => Promise<PlaylistBrowserMutationResult>;
@@ -418,6 +421,7 @@ export function createIpcDepsRuntime(options: IpcDepsRuntimeOptions): IpcService
         entries: [],
       })),
     appendClipboardVideoToQueue: options.appendClipboardVideoToQueue,
+    getChangelogSnapshot: options.getChangelogSnapshot,
     getPlaylistBrowserSnapshot: options.getPlaylistBrowserSnapshot,
     appendPlaylistBrowserFile: options.appendPlaylistBrowserFile,
     playPlaylistBrowserIndex: options.playPlaylistBrowserIndex,
@@ -820,6 +824,17 @@ export function registerIpcHandlers(deps: IpcServiceDeps, ipc: IpcMainRegistrar
     return deps.appendClipboardVideoToQueue();
   });
 
+  ipc.handle(IPC_CHANNELS.request.getChangelogSnapshot, async (_event, payload: unknown) => {
+    const refresh =
+      typeof payload === 'object' && payload !== null && 'refresh' in payload
+        ? (payload as { refresh?: unknown }).refresh === true
+        : false;
+    if (!deps.getChangelogSnapshot) {
+      throw new Error('Changelog service is unavailable.');
+    }
+    return await deps.getChangelogSnapshot({ refresh });
+  });
+
   ipc.handle(IPC_CHANNELS.request.getPlaylistBrowserSnapshot, async () => {
     return await deps.getPlaylistBrowserSnapshot();
   });

src/core/services/startup-bootstrap.test.ts

@@ -205,7 +205,7 @@ test('runStartupBootstrapRuntime skips lifecycle when generate-config flow handl
   assert.deepEqual(calls, ['setLog:warn:cli', 'forceX11', 'enforceWayland']);
 });
 
-test('runStartupBootstrapRuntime enables quiet background mode by default', () => {
+test('runStartupBootstrapRuntime lets config govern background log level by default', () => {
   const calls: string[] = [];
   const args = makeArgs({ background: true });
 
@@ -222,7 +222,7 @@ test('runStartupBootstrapRuntime enables quiet background mode by default', () =
   });
 
   assert.equal(result.backgroundMode, true);
-  assert.deepEqual(calls, ['setLog:warn:cli', 'forceX11', 'enforceWayland', 'startLifecycle']);
+  assert.deepEqual(calls, ['forceX11', 'enforceWayland', 'startLifecycle']);
 });
 
 test('runStartupBootstrapRuntime enables quiet update mode by default', () => {

src/core/services/startup.ts

@@ -45,7 +45,7 @@ export function runStartupBootstrapRuntime(
 
   if (initialArgs.logLevel) {
     deps.setLogLevel(initialArgs.logLevel, 'cli');
-  } else if (initialArgs.background || initialArgs.update) {
+  } else if (initialArgs.update) {
     deps.setLogLevel('warn', 'cli');
   }
 

src/core/services/stats-server/library-routes.ts

@@ -1,5 +1,6 @@
 import type { Hono } from 'hono';
 import { statsJson } from '../../../types/stats-http-contract.js';
+import { UNKNOWN_MOVE_TARGET_MESSAGE } from '../immersion-tracker/anime-merge.js';
 import type { ImmersionTrackerService } from '../immersion-tracker-service.js';
 import {
   buildSentenceSearchOptions,
@@ -7,6 +8,7 @@ import {
   parseBooleanQuery,
   parseExcludedWordsBody,
   parseIntQuery,
+  parsePositiveIdList,
 } from './route-support.js';
 
 export function registerStatsLibraryRoutes(
@@ -130,6 +132,19 @@ export function registerStatsLibraryRoutes(
     return c.json(statsJson('animeLibrary', rows));
   });
 
+  app.get('/api/stats/anime/merge-recommendations', async (c) => {
+    const recommendations = await tracker.getAnimeMergeRecommendations();
+    return c.json(statsJson('animeMergeRecommendations', { recommendations }));
+  });
+
+  app.delete('/api/stats/anime/merge-recommendations/:recommendationId', async (c) => {
+    const recommendationId = parseIntQuery(c.req.param('recommendationId'), 0);
+    if (recommendationId <= 0) return c.body(null, 400);
+    const dismissed = await tracker.dismissAnimeMergeRecommendation(recommendationId);
+    if (!dismissed) return c.body(null, 404);
+    return c.json(statsJson('dismissAnimeMergeRecommendation', { ok: true }));
+  });
+
   app.get('/api/stats/anime/:animeId', async (c) => {
     const animeId = parseIntQuery(c.req.param('animeId'), 0);
     if (animeId <= 0) return c.body(null, 400);
@@ -197,4 +212,50 @@ export function registerStatsLibraryRoutes(
     await tracker.deleteAnime(animeId);
     return c.json(statsJson('deleteAnime', { ok: true }));
   });
+
+  app.post('/api/stats/anime/:animeId/merge', async (c) => {
+    const animeId = parseIntQuery(c.req.param('animeId'), 0);
+    if (animeId <= 0) return c.body(null, 400);
+    const body = await c.req.json().catch(() => null);
+    const sourceAnimeIds = parsePositiveIdList(body?.sourceAnimeIds).filter((id) => id !== animeId);
+    if (sourceAnimeIds.length === 0) return c.body(null, 400);
+    const summary = await tracker.mergeAnime(animeId, sourceAnimeIds);
+    // Nothing folded means the target or every source was already gone, so the
+    // caller should not be told the merge succeeded.
+    if (summary.mergedAnimeIds.length === 0) return c.body(null, 404);
+    return c.json(
+      statsJson('mergeAnime', {
+        ok: true,
+        animeId: summary.survivingAnimeId,
+        mergedAnimeIds: summary.mergedAnimeIds,
+        movedVideos: summary.movedVideos,
+      }),
+    );
+  });
+
+  app.patch('/api/stats/media/:videoId/anime', async (c) => {
+    const videoId = parseIntQuery(c.req.param('videoId'), 0);
+    if (videoId <= 0) return c.body(null, 400);
+    const body = await c.req.json().catch(() => null);
+    const animeId = Number.isSafeInteger(body?.animeId) ? (body.animeId as number) : 0;
+    if (animeId <= 0) return c.body(null, 400);
+    try {
+      const summary = await tracker.moveVideoToAnime(videoId, animeId);
+      return c.json(
+        statsJson('moveVideoToAnime', {
+          ok: true,
+          animeId: summary.targetAnimeId,
+          previousAnimeId: summary.previousAnimeId,
+          removedPreviousAnime: summary.removedPreviousAnime,
+        }),
+      );
+    } catch (error) {
+      // Only a missing episode or entry is a 404; storage failures must not be
+      // reported to the caller as "not found".
+      if (error instanceof Error && error.message === UNKNOWN_MOVE_TARGET_MESSAGE) {
+        return c.body(null, 404);
+      }
+      throw error;
+    }
+  });
 }

src/core/services/stats-server/mining-routes.ts

@@ -11,7 +11,7 @@ import {
   resolveSecondarySubtitleTextFromSidecar,
 } from '../secondary-subtitle-sidecar.js';
 import {
-  applyStatsWordAndSentenceCardFields,
+  applyStatsWordCardFields,
   createStatsMiningContext,
   getStatsDirectMiningAudioFieldNames,
   getStatsWordMiningAudioFieldName,
@@ -272,7 +272,7 @@ export function registerStatsMiningRoutes(app: Hono, options?: StatsMiningRouteO
       const imageFieldName = ankiConfig.fields?.image ?? 'Picture';
 
       mediaFields[sentenceFieldName] = highlightedSentence;
-      applyStatsWordAndSentenceCardFields(mediaFields, noteInfo, ankiConfig);
+      applyStatsWordCardFields(mediaFields, noteInfo, ankiConfig);
 
       if (audioBuffer) {
         const audioFilename = `subminer_audio_${timestamp}_${noteId}.mp3`;

src/core/services/stats-server/mining-support.ts

@@ -1,5 +1,7 @@
 import type { MediaGenerator } from '../../../media-generator.js';
 import type { AnkiConnectConfig } from '../../../types.js';
+import { applyCardKindFlagFields } from '../../../anki-integration/card-kinds.js';
+import { resolveWordCardKind } from '../../../anki-integration/note-field-utils.js';
 import { createLogger } from '../../../logger.js';
 import type { RetimedSecondarySubtitleInput } from '../secondary-subtitle-sidecar.js';
 
@@ -94,20 +96,22 @@ export function shouldUseStatsLapisKikuCardFields(ankiConfig: AnkiConnectConfig)
   return ankiConfig.isLapis?.enabled === true || ankiConfig.isKiku?.enabled === true;
 }
 
-export function applyStatsWordAndSentenceCardFields(
+export function applyStatsWordCardFields(
   fields: Record<string, string>,
   noteInfo: StatsServerNoteInfo | null,
   ankiConfig: AnkiConnectConfig,
 ): void {
-  if (!shouldUseStatsLapisKikuCardFields(ankiConfig) || !noteInfo) return;
+  if (!noteInfo) return;
-  const wordAndSentenceFlag = resolveStatsNoteFieldName(noteInfo, 'IsWordAndSentenceCard');
+  const cardKind = resolveWordCardKind(noteInfo, {
-  if (!wordAndSentenceFlag) return;
+    lapisEnabled: ankiConfig.isLapis?.enabled === true,
+    kikuEnabled: ankiConfig.isKiku?.enabled === true,
+    wordCardKind: ankiConfig.lapisKiku?.wordCardKind,
+  });
+  if (!cardKind) return;
 
-  fields[wordAndSentenceFlag] = 'x';
+  applyCardKindFlagFields(fields, cardKind, (preferredName) =>
-  for (const flagName of ['IsSentenceCard', 'IsAudioCard']) {
+    resolveStatsNoteFieldName(noteInfo, preferredName),
-    const resolved = resolveStatsNoteFieldName(noteInfo, flagName);
+  );
-    if (resolved && resolved !== wordAndSentenceFlag) fields[resolved] = '';
-  }
 }
 
 export function getStatsDirectMiningAudioFieldNames(

src/core/services/stats-server/route-support.ts

@@ -170,6 +170,18 @@ export async function enrichSessionsWithKnownWordMetrics<
   );
 }
 
+/** Deduplicated positive integer ids from an untrusted JSON body field. */
+export function parsePositiveIdList(raw: unknown): number[] {
+  if (!Array.isArray(raw)) return [];
+  const ids = new Set<number>();
+  for (const value of raw) {
+    if (Number.isSafeInteger(value) && (value as number) > 0) {
+      ids.add(value as number);
+    }
+  }
+  return [...ids];
+}
+
 export function parseBooleanQuery(raw: string | undefined, fallback: boolean): boolean {
   if (raw === undefined) return fallback;
   const normalized = raw.trim().toLowerCase();

src/core/services/stats-sync/merge-catalog.ts

@@ -28,6 +28,7 @@ const VIDEO_COPY_COLUMNS = [
   'parser_source',
   'parser_confidence',
   'parse_metadata_json',
+  'anime_assignment_locked',
   'watched',
   'duration_ms',
   'file_size_bytes',

src/core/services/subsync.test.ts

@@ -8,6 +8,7 @@ import {
   runSubsyncManual,
   triggerSubsyncFromConfig,
 } from './subsync';
+import type { SubsyncManualPayload } from '../../types';
 
 function makeDeps(
   overrides: Partial<TriggerSubsyncFromConfigDeps> = {},
@@ -76,7 +77,7 @@ test('triggerSubsyncFromConfig opens manual picker', async () => {
   await triggerSubsyncFromConfig(
     makeDeps({
       openManualPicker: (payload) => {
-        payloadTrackCount = payload.sourceTracks.length;
+        payloadTrackCount = payload.subtitleTracks.length;
         ffsubsyncAvailable = payload.ffsubsyncAvailable;
       },
       showMpvOsd: (text) => {
@@ -88,9 +89,9 @@ test('triggerSubsyncFromConfig opens manual picker', async () => {
     }),
   );
 
-  assert.equal(payloadTrackCount, 1);
+  assert.equal(payloadTrackCount, 2);
   assert.equal(ffsubsyncAvailable, true);
-  assert.ok(osd.includes('Subsync: choose engine and source'));
+  assert.ok(osd.includes('Subsync: choose engine and subtitles'));
   assert.equal(inProgressState, false);
 });
 
@@ -140,7 +141,7 @@ test('triggerSubsyncFromConfig does not run automatic sync', async () => {
   await triggerSubsyncFromConfig(
     makeDeps({
       openManualPicker: (payload) => {
-        payloadTrackCount = payload.sourceTracks.length;
+        payloadTrackCount = payload.subtitleTracks.length;
       },
       showMpvOsd: (text) => {
         osd.push(text);
@@ -152,9 +153,9 @@ test('triggerSubsyncFromConfig does not run automatic sync', async () => {
     }),
   );
 
-  assert.equal(payloadTrackCount, 1);
+  assert.equal(payloadTrackCount, 2);
   assert.equal(spinnerRan, false);
-  assert.deepEqual(osd, ['Subsync: choose engine and source']);
+  assert.deepEqual(osd, ['Subsync: choose engine and subtitles']);
 });
 
 test('triggerSubsyncFromConfig dedupes repeated subtitle source tracks', async () => {
@@ -195,12 +196,71 @@ test('triggerSubsyncFromConfig dedupes repeated subtitle source tracks', async (
         },
       }),
       openManualPicker: (payload) => {
-        payloadTrackCount = payload.sourceTracks.length;
+        payloadTrackCount = payload.subtitleTracks.length;
       },
     }),
   );
 
-  assert.equal(payloadTrackCount, 1);
+  assert.equal(payloadTrackCount, 2);
+});
+
+test('triggerSubsyncFromConfig keeps both active tracks when they share a file', async () => {
+  let payload: SubsyncManualPayload | null = null;
+
+  await triggerSubsyncFromConfig(
+    makeDeps({
+      getMpvClient: () => ({
+        connected: true,
+        currentAudioStreamIndex: null,
+        send: () => {},
+        requestProperty: async (name: string) => {
+          if (name === 'path') return '/tmp/video.mkv';
+          if (name === 'sid') return 1;
+          if (name === 'secondary-sid') return 2;
+          if (name === 'track-list') {
+            // mpv appends a duplicate entry when the same file is re-added, so
+            // the primary and secondary slots can point at one path.
+            return [
+              {
+                id: 1,
+                type: 'sub',
+                selected: true,
+                external: true,
+                'external-filename': '/tmp/ref.srt',
+              },
+              {
+                id: 2,
+                type: 'sub',
+                selected: true,
+                external: true,
+                'external-filename': '/tmp/ref.srt',
+              },
+              {
+                id: 3,
+                type: 'sub',
+                selected: false,
+                external: true,
+                'external-filename': '/tmp/ref.srt',
+              },
+            ];
+          }
+          return null;
+        },
+      }),
+      openManualPicker: (nextPayload) => {
+        payload = nextPayload;
+      },
+    }),
+  );
+
+  assert.ok(payload);
+  const resolved = payload as SubsyncManualPayload;
+  assert.deepEqual(
+    resolved.subtitleTracks.map((track) => track.id),
+    [1, 2],
+  );
+  assert.equal(resolved.defaultReferenceTrackId, 2);
+  assert.equal(resolved.defaultTargetTrackId, 1);
 });
 
 test('triggerSubsyncFromConfig reports failures to OSD', async () => {
@@ -217,15 +277,157 @@ test('triggerSubsyncFromConfig reports failures to OSD', async () => {
   assert.ok(osd.some((line) => line.startsWith('Subsync failed: MPV not connected')));
 });
 
-test('runSubsyncManual requires a source track for alass', async () => {
+test('runSubsyncManual requires a reference track for alass', async () => {
-  const result = await runSubsyncManual({ engine: 'alass', sourceTrackId: null }, makeDeps());
+  const result = await runSubsyncManual({ engine: 'alass', referenceTrackId: null }, makeDeps());
 
   assert.deepEqual(result, {
     ok: false,
-    message: 'Select a subtitle source track for alass',
+    message: 'Select a reference subtitle track for alass',
   });
 });
 
+test('runSubsyncManual rejects alass when reference and target are the same track', async () => {
+  const result = await runSubsyncManual(
+    { engine: 'alass', referenceTrackId: 2, targetTrackId: 2 },
+    makeDeps(),
+  );
+
+  assert.deepEqual(result, {
+    ok: false,
+    message: 'Reference and out-of-sync subtitles must be different tracks',
+  });
+});
+
+test('runSubsyncManual rejects an unknown target track', async () => {
+  const result = await runSubsyncManual(
+    { engine: 'alass', referenceTrackId: 2, targetTrackId: 99 },
+    makeDeps(),
+  );
+
+  assert.deepEqual(result, {
+    ok: false,
+    message: 'Select the out-of-sync subtitle track to retime',
+  });
+});
+
+test('runSubsyncManual rejects the video reference for remote media', async () => {
+  const result = await runSubsyncManual(
+    { engine: 'alass', referenceMode: 'video' },
+    makeDeps({
+      getMpvClient: () => ({
+        connected: true,
+        currentAudioStreamIndex: null,
+        send: () => {},
+        requestProperty: async (name: string) => {
+          if (name === 'path') return 'https://jellyfin.example/Videos/movie/stream.mkv';
+          if (name === 'sid') return 1;
+          if (name === 'secondary-sid') return null;
+          if (name === 'track-list') {
+            return [{ id: 1, type: 'sub', selected: true, lang: 'jpn' }];
+          }
+          return null;
+        },
+      }),
+    }),
+  );
+
+  assert.equal(result.ok, false);
+  assert.match(result.message, /cannot use a stream URL as reference/);
+});
+
+test('openSubsyncManualPicker defaults the reference to the secondary subtitle track', async () => {
+  let payload: SubsyncManualPayload | null = null;
+
+  await triggerSubsyncFromConfig(
+    makeDeps({
+      getMpvClient: () => ({
+        connected: true,
+        currentAudioStreamIndex: null,
+        send: () => {},
+        requestProperty: async (name: string) => {
+          if (name === 'path') return '/tmp/video.mkv';
+          if (name === 'sid') return 1;
+          if (name === 'secondary-sid') return 3;
+          if (name === 'track-list') {
+            return [
+              { id: 1, type: 'sub', selected: true, lang: 'jpn' },
+              {
+                id: 2,
+                type: 'sub',
+                selected: false,
+                external: true,
+                lang: 'eng',
+                'external-filename': '/tmp/other.srt',
+              },
+              {
+                id: 3,
+                type: 'sub',
+                selected: true,
+                external: true,
+                lang: 'eng',
+                'external-filename': '/tmp/secondary.srt',
+              },
+            ];
+          }
+          return null;
+        },
+      }),
+      openManualPicker: (nextPayload) => {
+        payload = nextPayload;
+      },
+    }),
+  );
+
+  assert.ok(payload);
+  const resolved = payload as SubsyncManualPayload;
+  assert.deepEqual(
+    resolved.subtitleTracks.map((track) => track.id),
+    [1, 2, 3],
+  );
+  assert.equal(resolved.defaultReferenceTrackId, 3);
+  assert.equal(resolved.defaultTargetTrackId, 1);
+  assert.equal(resolved.videoReferenceAvailable, true);
+});
+
+test('openSubsyncManualPicker never defaults to a reference missing from the track list', async () => {
+  let payload: SubsyncManualPayload | null = null;
+
+  await triggerSubsyncFromConfig(
+    makeDeps({
+      getMpvClient: () => ({
+        connected: true,
+        currentAudioStreamIndex: null,
+        send: () => {},
+        requestProperty: async (name: string) => {
+          if (name === 'path') return '/tmp/video.mkv';
+          if (name === 'sid') return 1;
+          if (name === 'secondary-sid') return 2;
+          if (name === 'track-list') {
+            return [
+              { id: 1, type: 'sub', selected: true, lang: 'jpn' },
+              // Secondary track with no usable file path: filtered out of the picker.
+              { id: 2, type: 'sub', selected: true, external: true, 'external-filename': '' },
+              { id: 3, type: 'sub', selected: false, lang: 'eng' },
+            ];
+          }
+          return null;
+        },
+      }),
+      openManualPicker: (nextPayload) => {
+        payload = nextPayload;
+      },
+    }),
+  );
+
+  assert.ok(payload);
+  const resolved = payload as SubsyncManualPayload;
+  assert.deepEqual(
+    resolved.subtitleTracks.map((track) => track.id),
+    [1, 3],
+  );
+  assert.equal(resolved.defaultReferenceTrackId, 3);
+});
+
 test('triggerSubsyncFromConfig does not validate sync tool paths before manual selection', async () => {
   const osd: string[] = [];
   const inProgress: boolean[] = [];
@@ -242,7 +444,7 @@ test('triggerSubsyncFromConfig does not validate sync tool paths before manual s
         inProgress.push(value);
       },
       openManualPicker: (payload) => {
-        payloadTrackCount = payload.sourceTracks.length;
+        payloadTrackCount = payload.subtitleTracks.length;
       },
       showMpvOsd: (text) => {
         osd.push(text);
@@ -251,8 +453,8 @@ test('triggerSubsyncFromConfig does not validate sync tool paths before manual s
   );
 
   assert.deepEqual(inProgress, [false]);
-  assert.equal(payloadTrackCount, 1);
+  assert.equal(payloadTrackCount, 2);
-  assert.deepEqual(osd, ['Subsync: choose engine and source']);
+  assert.deepEqual(osd, ['Subsync: choose engine and subtitles']);
 });
 
 function writeExecutableScript(filePath: string, content: string): void {
@@ -333,7 +535,7 @@ test('runSubsyncManual constructs ffsubsync command and returns success', async
     }),
   });
 
-  const result = await runSubsyncManual({ engine: 'ffsubsync', sourceTrackId: null }, deps);
+  const result = await runSubsyncManual({ engine: 'ffsubsync' }, deps);
 
   assert.equal(result.ok, true);
   assert.equal(result.message, 'Subtitle synchronized with ffsubsync');
@@ -346,7 +548,7 @@ test('runSubsyncManual constructs ffsubsync command and returns success', async
   const ffOutputFlagIndex = ffArgs.indexOf('-o');
   assert.equal(ffOutputFlagIndex >= 0, true);
   assert.equal(ffArgs[ffOutputFlagIndex + 1], toShellPath(primaryPath));
-  assert.equal(sentCommands[0]?.[0], 'sub_add');
+  assert.equal(sentCommands[0]?.[0], 'sub-add');
   assert.deepEqual(sentCommands[1], ['set_property', 'sub-delay', 0]);
 });
 
@@ -399,7 +601,7 @@ test('runSubsyncManual writes deterministic _retimed filename when replace is fa
     }),
   });
 
-  const result = await runSubsyncManual({ engine: 'ffsubsync', sourceTrackId: null }, deps);
+  const result = await runSubsyncManual({ engine: 'ffsubsync' }, deps);
 
   assert.equal(result.ok, true);
   const ffArgs = fs.readFileSync(ffsubsyncLogPath, 'utf8').trim().split('\n');
@@ -453,7 +655,7 @@ test('runSubsyncManual reports ffsubsync command failures with details', async (
     }),
   });
 
-  const result = await runSubsyncManual({ engine: 'ffsubsync', sourceTrackId: null }, deps);
+  const result = await runSubsyncManual({ engine: 'ffsubsync' }, deps);
 
   assert.equal(result.ok, false);
   assert.equal(result.message.startsWith('ffsubsync synchronization failed'), true);
@@ -518,7 +720,7 @@ test('runSubsyncManual constructs alass command and returns failure on non-zero
     }),
   });
 
-  const result = await runSubsyncManual({ engine: 'alass', sourceTrackId: 2 }, deps);
+  const result = await runSubsyncManual({ engine: 'alass', referenceTrackId: 2 }, deps);
 
   assert.equal(result.ok, false);
   assert.equal(typeof result.message, 'string');
@@ -528,6 +730,179 @@ test('runSubsyncManual constructs alass command and returns failure on non-zero
   assert.equal(alassArgs[1], toShellPath(primaryPath));
 });
 
+function makeAlassSelectionDeps(tmpDir: string): {
+  deps: TriggerSubsyncFromConfigDeps;
+  alassLogPath: string;
+  videoPath: string;
+  primaryPath: string;
+  sourcePath: string;
+  sentCommands: Array<Array<string | number>>;
+} {
+  const alassLogPath = path.join(tmpDir, 'alass-args.log');
+  const alassPath = path.join(tmpDir, 'alass.sh');
+  const ffmpegPath = path.join(tmpDir, 'ffmpeg.sh');
+  const ffsubsyncPath = path.join(tmpDir, 'ffsubsync.sh');
+  const videoPath = path.join(tmpDir, 'video.mkv');
+  const primaryPath = path.join(tmpDir, 'primary.srt');
+  const sourcePath = path.join(tmpDir, 'source.srt');
+
+  fs.writeFileSync(videoPath, 'video');
+  fs.writeFileSync(primaryPath, 'sub');
+  fs.writeFileSync(sourcePath, 'sub2');
+  writeExecutableScript(ffmpegPath, '#!/bin/sh\nexit 0\n');
+  writeExecutableScript(ffsubsyncPath, '#!/bin/sh\nexit 0\n');
+  writeExecutableScript(
+    alassPath,
+    `#!/bin/sh\n: > "${toShellPath(alassLogPath)}"\nfor arg in "$@"; do printf '%s\\n' "$arg" >> "${toShellPath(alassLogPath)}"; done\n: > "$3"\nexit 0\n`,
+  );
+
+  const trackList: Array<Record<string, unknown>> = [
+    { id: 1, type: 'sub', selected: true, external: true, 'external-filename': primaryPath },
+    { id: 2, type: 'sub', selected: true, external: true, 'external-filename': sourcePath },
+  ];
+  const sentCommands: Array<Array<string | number>> = [];
+  const deps = makeDeps({
+    getMpvClient: () => ({
+      connected: true,
+      currentAudioStreamIndex: null,
+      send: (payload) => {
+        sentCommands.push(payload.command);
+        if (payload.command[0] === 'sub-add' || payload.command[0] === 'sub_add') {
+          trackList.push({
+            id: trackList.length + 1,
+            type: 'sub',
+            selected: false,
+            external: true,
+            'external-filename': payload.command[1],
+          });
+        }
+      },
+      requestProperty: async (name: string) => {
+        if (name === 'path') return videoPath;
+        if (name === 'sid') return 1;
+        if (name === 'secondary-sid') return 2;
+        if (name === 'track-list') return trackList;
+        return null;
+      },
+    }),
+    getResolvedConfig: () => ({
+      alassPath,
+      ffsubsyncPath,
+      ffmpegPath,
+      replace: false,
+    }),
+  });
+
+  return { deps, alassLogPath, videoPath, primaryPath, sourcePath, sentCommands };
+}
+
+test('runSubsyncManual uses the video file as alass reference when requested', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'subsync-alass-video-ref-'));
+  const { deps, alassLogPath, videoPath, primaryPath } = makeAlassSelectionDeps(tmpDir);
+
+  const result = await runSubsyncManual({ engine: 'alass', referenceMode: 'video' }, deps);
+
+  assert.equal(result.ok, true);
+  const alassArgs = fs.readFileSync(alassLogPath, 'utf8').trim().split('\n');
+  assert.equal(alassArgs[0], toShellPath(videoPath));
+  assert.equal(alassArgs[1], toShellPath(primaryPath));
+});
+
+test('runSubsyncManual retimes the selected target track instead of the primary', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'subsync-alass-target-'));
+  const { deps, alassLogPath, primaryPath, sourcePath, sentCommands } =
+    makeAlassSelectionDeps(tmpDir);
+
+  const result = await runSubsyncManual(
+    { engine: 'alass', referenceTrackId: 1, targetTrackId: 2 },
+    deps,
+  );
+
+  assert.equal(result.ok, true);
+  const alassArgs = fs.readFileSync(alassLogPath, 'utf8').trim().split('\n');
+  assert.equal(alassArgs[0], toShellPath(primaryPath));
+  assert.equal(alassArgs[1], toShellPath(sourcePath));
+  assert.equal(sentCommands[0]?.[0], 'sub-add');
+  assert.equal(sentCommands[0]?.[2], 'auto');
+});
+
+test('runSubsyncManual keeps a retimed secondary track in the secondary slot', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'subsync-alass-secondary-slot-'));
+  const alassPath = path.join(tmpDir, 'alass.sh');
+  const ffmpegPath = path.join(tmpDir, 'ffmpeg.sh');
+  const ffsubsyncPath = path.join(tmpDir, 'ffsubsync.sh');
+  const videoPath = path.join(tmpDir, 'video.mkv');
+  const primaryPath = path.join(tmpDir, 'ja.srt');
+  const secondaryPath = path.join(tmpDir, 'en.srt');
+  const retimedPath = path.join(tmpDir, 'en_retimed.srt');
+
+  fs.writeFileSync(videoPath, 'video');
+  fs.writeFileSync(primaryPath, 'ja');
+  fs.writeFileSync(secondaryPath, 'en');
+  writeExecutableScript(ffmpegPath, '#!/bin/sh\nexit 0\n');
+  writeExecutableScript(ffsubsyncPath, '#!/bin/sh\nexit 0\n');
+  writeExecutableScript(alassPath, '#!/bin/sh\n: > "$3"\nexit 0\n');
+
+  const trackList: Array<Record<string, unknown>> = [
+    { id: 1, type: 'sub', selected: true, external: true, 'external-filename': primaryPath },
+    { id: 2, type: 'sub', selected: true, external: true, 'external-filename': secondaryPath },
+  ];
+  const sentCommands: Array<Array<string | number>> = [];
+  const deps = makeDeps({
+    getMpvClient: () => ({
+      connected: true,
+      currentAudioStreamIndex: null,
+      send: (payload) => {
+        sentCommands.push(payload.command);
+        if (payload.command[0] === 'sub-add') {
+          trackList.push({
+            id: 3,
+            type: 'sub',
+            selected: false,
+            external: true,
+            'external-filename': payload.command[1],
+          });
+        }
+      },
+      requestProperty: async (name: string) => {
+        if (name === 'path') return videoPath;
+        if (name === 'sid') return 1;
+        if (name === 'secondary-sid') return 2;
+        if (name === 'track-list') return trackList;
+        return null;
+      },
+    }),
+    getResolvedConfig: () => ({
+      alassPath,
+      ffsubsyncPath,
+      ffmpegPath,
+      replace: false,
+    }),
+  });
+
+  const result = await runSubsyncManual(
+    { engine: 'alass', referenceTrackId: 1, targetTrackId: 2 },
+    deps,
+  );
+
+  assert.equal(result.ok, true);
+  assert.deepEqual(sentCommands[0], ['sub-add', retimedPath, 'auto']);
+  assert.deepEqual(sentCommands[1], ['set_property', 'secondary-sub-delay', 0]);
+  assert.deepEqual(sentCommands[2], ['set_property', 'secondary-sid', 3]);
+  assert.equal(
+    sentCommands.some((command) => command[1] === 'sub-delay'),
+    false,
+  );
+  assert.equal(
+    sentCommands.some((command) => command[1] === 'sid'),
+    false,
+  );
+  assert.equal(
+    sentCommands.some((command) => command[1] === 'sid'),
+    false,
+  );
+});
+
 test('runSubsyncManual keeps internal alass source file alive until sync finishes', async () => {
   const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'subsync-alass-internal-source-'));
   const alassPath = path.join(tmpDir, 'alass.sh');
@@ -589,11 +964,11 @@ test('runSubsyncManual keeps internal alass source file alive until sync finishe
     }),
   });
 
-  const result = await runSubsyncManual({ engine: 'alass', sourceTrackId: 2 }, deps);
+  const result = await runSubsyncManual({ engine: 'alass', referenceTrackId: 2 }, deps);
 
   assert.equal(result.ok, true);
   assert.equal(result.message, 'Subtitle synchronized with alass');
-  assert.equal(sentCommands[0]?.[0], 'sub_add');
+  assert.equal(sentCommands[0]?.[0], 'sub-add');
   assert.deepEqual(sentCommands[1], ['set_property', 'sub-delay', 0]);
 });
 
@@ -645,7 +1020,7 @@ test('runSubsyncManual resolves string sid values from mpv stream properties', a
     }),
   });
 
-  const result = await runSubsyncManual({ engine: 'ffsubsync', sourceTrackId: null }, deps);
+  const result = await runSubsyncManual({ engine: 'ffsubsync' }, deps);
 
   assert.equal(result.ok, true);
   assert.equal(result.message, 'Subtitle synchronized with ffsubsync');

src/core/services/subsync.ts

@@ -21,6 +21,11 @@ interface FileExtractionResult {
   temporary: boolean;
 }
 
+type SubtitleSlot = 'primary' | 'secondary';
+
+const SYNCED_TRACK_LOOKUP_ATTEMPTS = 5;
+const SYNCED_TRACK_LOOKUP_RETRY_MS = 100;
+
 function summarizeCommandFailure(command: string, result: CommandResult): string {
   const parts = [
     `code=${result.code ?? 'n/a'}`,
@@ -90,16 +95,28 @@ function getSourceTrackIdentity(track: MpvTrack): string {
   return 'unknown';
 }
 
-function dedupeSourceTracks(tracks: MpvTrack[]): MpvTrack[] {
+function isPinned(track: MpvTrack, pinnedIds: Set<number>): boolean {
-  const deduped = new Map<string, MpvTrack>();
+  return typeof track.id === 'number' && pinnedIds.has(track.id);
+}
+
+// Pinned tracks (the active primary/secondary) always survive, even when two of
+// them point at the same file; only unpinned duplicates are collapsed.
+function dedupeSubtitleTracks(tracks: MpvTrack[], pinnedIds: Set<number>): MpvTrack[] {
+  const pinnedIdentities = new Set(
+    tracks.filter((track) => isPinned(track, pinnedIds)).map(getSourceTrackIdentity),
+  );
+  const winners = new Map<string, MpvTrack>();
   for (const track of tracks) {
+    if (isPinned(track, pinnedIds)) continue;
     const identity = getSourceTrackIdentity(track);
-    const existing = deduped.get(identity);
+    if (pinnedIdentities.has(identity)) continue;
+    const existing = winners.get(identity);
     if (!existing || (track.selected && !existing.selected)) {
-      deduped.set(identity, track);
+      winners.set(identity, track);
     }
   }
-  return [...deduped.values()];
+  const kept = new Set(winners.values());
+  return tracks.filter((track) => isPinned(track, pinnedIds) || kept.has(track));
 }
 
 export interface TriggerSubsyncFromConfigDeps extends SubsyncCoreDeps {
@@ -142,20 +159,21 @@ async function gatherSubsyncContext(client: MpvClientLike): Promise<SubsyncConte
   }
 
   const secondaryTrack = subtitleTracks.find((track) => track.id === secondarySid) ?? null;
-  const sourceTracks = subtitleTracks
+  const usableTracks = subtitleTracks.filter((track) => {
-    .filter((track) => track.id !== sid)
+    if (typeof track.id !== 'number') return false;
-    .filter((track) => {
+    if (!track.external) return true;
-      if (!track.external) return true;
+    const filename = track['external-filename'];
-      const filename = track['external-filename'];
+    return typeof filename === 'string' && filename.length > 0;
-      return typeof filename === 'string' && filename.length > 0;
+  });
-    });
-  const uniqueSourceTracks = dedupeSourceTracks(sourceTracks);
 
   return {
     videoPath,
     primaryTrack,
     secondaryTrack,
-    sourceTracks: uniqueSourceTracks,
+    subtitleTracks: dedupeSubtitleTracks(
+      usableTracks,
+      new Set([sid, secondarySid].filter((id): id is number => typeof id === 'number')),
+    ),
     audioStreamIndex: client.currentAudioStreamIndex,
   };
 }
@@ -271,41 +289,104 @@ async function runFfsubsyncSync(
   return runCommand(ffsubsyncPath, args);
 }
 
-function loadSyncedSubtitle(client: MpvClientLike, pathToLoad: string): void {
+function delay(ms: number): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+// mpv may echo the path back with different separators, and Windows paths are
+// case-insensitive, so compare normalized forms instead of raw strings.
+function normalizeSubtitlePathForCompare(value: string): string {
+  const normalized = value.replace(/\\/g, '/');
+  return process.platform === 'win32' ? normalized.toLowerCase() : normalized;
+}
+
+async function findAddedSubtitleTrackId(
+  client: MpvClientLike,
+  pathToLoad: string,
+): Promise<number | null> {
+  const wanted = normalizeSubtitlePathForCompare(pathToLoad);
+  // sub-add is queued, so the track may not appear in the first track-list reply.
+  for (let attempt = 0; attempt < SYNCED_TRACK_LOOKUP_ATTEMPTS; attempt += 1) {
+    let tracks: MpvTrack[] = [];
+    try {
+      const trackListRaw = await client.requestProperty('track-list');
+      tracks = Array.isArray(trackListRaw) ? normalizeTrackIds(trackListRaw as MpvTrack[]) : [];
+    } catch {
+      return null;
+    }
+    // Re-adding a file mpv already knows appends a duplicate entry; the newest
+    // one holds the retimed content, so prefer the last match.
+    const matches = tracks.filter((track) => {
+      if (track.type !== 'sub') return false;
+      const filename = track['external-filename'];
+      return typeof filename === 'string' && normalizeSubtitlePathForCompare(filename) === wanted;
+    });
+    const added = matches[matches.length - 1];
+    if (added && typeof added.id === 'number') {
+      return added.id;
+    }
+    if (attempt < SYNCED_TRACK_LOOKUP_ATTEMPTS - 1) {
+      await delay(SYNCED_TRACK_LOOKUP_RETRY_MS);
+    }
+  }
+  return null;
+}
+
+async function loadSyncedSubtitle(
+  client: MpvClientLike,
+  pathToLoad: string,
+  slot: SubtitleSlot,
+): Promise<void> {
   if (!client.connected) {
     throw new Error('MPV disconnected while loading subtitle');
   }
-  client.send({ command: ['sub_add', pathToLoad] });
+
+  if (slot === 'secondary') {
+    // Keep the primary track untouched: load without selecting, then point
+    // secondary-sid at the freshly added track.
+    client.send({ command: ['sub-add', pathToLoad, 'auto'] });
+    client.send({ command: ['set_property', 'secondary-sub-delay', 0] });
+    const addedTrackId = await findAddedSubtitleTrackId(client, pathToLoad);
+    if (addedTrackId === null) {
+      throw new Error('Synchronized subtitle did not appear in the mpv track list');
+    }
+    client.send({ command: ['set_property', 'secondary-sid', addedTrackId] });
+    return;
+  }
+
+  client.send({ command: ['sub-add', pathToLoad] });
   client.send({ command: ['set_property', 'sub-delay', 0] });
 }
 
 async function subsyncToReference(
   engine: 'alass' | 'ffsubsync',
   referenceFilePath: string,
+  targetTrack: MpvTrack,
   context: SubsyncContext,
   resolved: SubsyncResolvedConfig,
   client: MpvClientLike,
+  slot: SubtitleSlot,
 ): Promise<SubsyncResult> {
   const ffmpegPath = ensureExecutablePath(resolved.ffmpegPath, 'ffmpeg');
-  const primaryExtraction = await extractSubtitleTrackToFile(
+  const targetExtraction = await extractSubtitleTrackToFile(
     ffmpegPath,
     context.videoPath,
-    context.primaryTrack,
+    targetTrack,
   );
-  const replacePrimary = resolved.replace !== false && !primaryExtraction.temporary;
+  const replaceTarget = resolved.replace !== false && !targetExtraction.temporary;
-  const outputPath = buildRetimedPath(primaryExtraction.path, replacePrimary);
+  const outputPath = buildRetimedPath(targetExtraction.path, replaceTarget);
 
   try {
     let result: CommandResult;
     if (engine === 'alass') {
       const alassPath = ensureExecutablePath(resolved.alassPath, 'alass');
-      result = await runAlassSync(alassPath, referenceFilePath, primaryExtraction.path, outputPath);
+      result = await runAlassSync(alassPath, referenceFilePath, targetExtraction.path, outputPath);
     } else {
       const ffsubsyncPath = ensureExecutablePath(resolved.ffsubsyncPath, 'ffsubsync');
       result = await runFfsubsyncSync(
         ffsubsyncPath,
         context.videoPath,
-        primaryExtraction.path,
+        targetExtraction.path,
         outputPath,
         context.audioStreamIndex,
       );
@@ -319,13 +400,13 @@ async function subsyncToReference(
       };
     }
 
-    loadSyncedSubtitle(client, outputPath);
+    await loadSyncedSubtitle(client, outputPath, slot);
     return {
       ok: true,
       message: `Subtitle synchronized with ${engine}`,
     };
   } finally {
-    cleanupTemporaryFile(primaryExtraction);
+    cleanupTemporaryFile(targetExtraction);
   }
 }
 
@@ -337,6 +418,25 @@ function validateFfsubsyncReference(videoPath: string): void {
   }
 }
 
+function resolveTargetTrack(
+  request: SubsyncManualRunRequest,
+  context: SubsyncContext,
+): MpvTrack | null {
+  if (request.targetTrackId === undefined || request.targetTrackId === null) {
+    return context.primaryTrack;
+  }
+  return getTrackById(context.subtitleTracks, request.targetTrackId);
+}
+
+// Retiming the secondary track must not steal the primary slot: the synced file
+// goes back where the out-of-sync one was.
+function resolveTargetSlot(targetTrack: MpvTrack, context: SubsyncContext): SubtitleSlot {
+  if (typeof targetTrack.id !== 'number') return 'primary';
+  if (targetTrack.id === context.primaryTrack.id) return 'primary';
+  if (context.secondaryTrack && targetTrack.id === context.secondaryTrack.id) return 'secondary';
+  return 'primary';
+}
+
 export async function runSubsyncManual(
   request: SubsyncManualRunRequest,
   deps: SubsyncCoreDeps,
@@ -345,6 +445,12 @@ export async function runSubsyncManual(
   const context = await gatherSubsyncContext(client);
   const resolved = deps.getResolvedConfig();
 
+  const targetTrack = resolveTargetTrack(request, context);
+  if (!targetTrack) {
+    return { ok: false, message: 'Select the out-of-sync subtitle track to retime' };
+  }
+  const targetSlot = resolveTargetSlot(targetTrack, context);
+
   if (request.engine === 'ffsubsync') {
     try {
       validateFfsubsyncReference(context.videoPath);
@@ -354,22 +460,64 @@ export async function runSubsyncManual(
         message: `ffsubsync synchronization failed: ${(error as Error).message}`,
       };
     }
-    return subsyncToReference('ffsubsync', context.videoPath, context, resolved, client);
+    return subsyncToReference(
+      'ffsubsync',
+      context.videoPath,
+      targetTrack,
+      context,
+      resolved,
+      client,
+      targetSlot,
+    );
   }
 
-  const sourceTrack = getTrackById(context.sourceTracks, request.sourceTrackId ?? null);
+  if (request.referenceMode === 'video') {
-  if (!sourceTrack) {
+    if (isRemoteMediaPath(context.videoPath)) {
-    return { ok: false, message: 'Select a subtitle source track for alass' };
+      return {
+        ok: false,
+        message:
+          'alass cannot use a stream URL as reference. Pick a reference subtitle track instead.',
+      };
+    }
+    return subsyncToReference(
+      'alass',
+      context.videoPath,
+      targetTrack,
+      context,
+      resolved,
+      client,
+      targetSlot,
+    );
+  }
+
+  const referenceTrack = getTrackById(context.subtitleTracks, request.referenceTrackId ?? null);
+  if (!referenceTrack) {
+    return { ok: false, message: 'Select a reference subtitle track for alass' };
+  }
+  if (referenceTrack.id === targetTrack.id) {
+    return { ok: false, message: 'Reference and out-of-sync subtitles must be different tracks' };
   }
 
   const ffmpegPath = ensureExecutablePath(resolved.ffmpegPath, 'ffmpeg');
-  let sourceExtraction: FileExtractionResult | null = null;
+  let referenceExtraction: FileExtractionResult | null = null;
   try {
-    sourceExtraction = await extractSubtitleTrackToFile(ffmpegPath, context.videoPath, sourceTrack);
+    referenceExtraction = await extractSubtitleTrackToFile(
-    return await subsyncToReference('alass', sourceExtraction.path, context, resolved, client);
+      ffmpegPath,
+      context.videoPath,
+      referenceTrack,
+    );
+    return await subsyncToReference(
+      'alass',
+      referenceExtraction.path,
+      targetTrack,
+      context,
+      resolved,
+      client,
+      targetSlot,
+    );
   } finally {
-    if (sourceExtraction) {
+    if (referenceExtraction) {
-      cleanupTemporaryFile(sourceExtraction);
+      cleanupTemporaryFile(referenceExtraction);
     }
   }
 }
@@ -377,14 +525,27 @@ export async function runSubsyncManual(
 export async function openSubsyncManualPicker(deps: TriggerSubsyncFromConfigDeps): Promise<void> {
   const client = getMpvClientForSubsync(deps);
   const context = await gatherSubsyncContext(client);
+  const subtitleTracks = context.subtitleTracks
+    .filter((track) => typeof track.id === 'number')
+    .map((track) => ({
+      id: track.id as number,
+      label: formatTrackLabel(track),
+    }));
+  const primaryTrackId =
+    typeof context.primaryTrack.id === 'number' ? context.primaryTrack.id : null;
+  const secondaryTrackId =
+    typeof context.secondaryTrack?.id === 'number' ? context.secondaryTrack.id : null;
   const payload: SubsyncManualPayload = {
+    subtitleTracks,
+    // The secondary track can be filtered or deduped out of the emitted list,
+    // so only default to it when the picker actually offers it.
+    defaultReferenceTrackId:
+      subtitleTracks.find((track) => track.id === secondaryTrackId)?.id ??
+      subtitleTracks.find((track) => track.id !== primaryTrackId)?.id ??
+      null,
+    defaultTargetTrackId: primaryTrackId,
+    videoReferenceAvailable: !isRemoteMediaPath(context.videoPath),
     ffsubsyncAvailable: !isRemoteMediaPath(context.videoPath),
-    sourceTracks: context.sourceTracks
-      .filter((track) => typeof track.id === 'number')
-      .map((track) => ({
-        id: track.id as number,
-        label: formatTrackLabel(track),
-      })),
   };
   deps.openManualPicker(payload);
 }
@@ -397,7 +558,7 @@ export async function triggerSubsyncFromConfig(deps: TriggerSubsyncFromConfigDep
 
   try {
     await openSubsyncManualPicker(deps);
-    deps.showMpvOsd('Subsync: choose engine and source');
+    deps.showMpvOsd('Subsync: choose engine and subtitles');
   } catch (error) {
     deps.showMpvOsd(`Subsync failed: ${(error as Error).message}`);
   } finally {

src/core/services/subtitle-cue-dedup.ts

@@ -0,0 +1,191 @@
+/*
+ * Duplicate/animation-burst collapsing for parsed subtitle cues.
+ *
+ * Split out of the cue parser so the parsing rules and the "is this run one animation?"
+ * heuristics can be read -- and tested -- on their own. The parser owns the cue shape;
+ * this module only decides which cues survive.
+ */
+
+import { hasAssTemporalOverride, isAnimatedAssEffectKind } from './ass-text';
+import type {
+  AnnotatedSubtitleCue,
+  SubtitleCue,
+  SubtitleSourceFormat,
+} from './subtitle-cue-parser';
+
+// Back-to-back frames of the same animation are authored flush against each other; a
+// tiny tolerance absorbs the centisecond rounding of the ASS timestamp format.
+const DUPLICATE_CUE_GAP_TOLERANCE_SECONDS = 0.05;
+// A burst is a *sequence*. Two adjacent events are two events, not an animation --
+// characters do repeat each other, and a repeated line can legitimately be short.
+const MIN_BURST_EVENTS = 3;
+// Real dialogue holds on screen for about a second, so a run with a couple of much
+// shorter events among them looks like frames. Used only alongside authoring evidence.
+const ANIMATION_FRAME_MAX_SECONDS = 0.3;
+// A karaoke run usually ends on a long "hold" frame, so not every event is short.
+const MIN_TAGGED_BURST_FRAMES = 2;
+// SRT and VTT carry no authoring metadata at all, so timing is the only signal available
+// -- which makes it the easiest one to get wrong. ASS->SRT conversion leaves frames at
+// ~0.04s, well under any real utterance, and a burst leaves many of them behind. Both
+// bounds are deliberately far stricter than the ASS path: a run of ordinary short lines
+// (`えっ` traded between characters) must not clear them.
+const TIMING_ONLY_FRAME_MAX_SECONDS = 0.1;
+const MIN_TIMING_ONLY_FRAMES = 5;
+
+function cueKey(cue: SubtitleCue): string {
+  return `${cue.startTime}|${cue.endTime}|${cue.text}`;
+}
+
+/**
+ * Identical text over an identical span is redundant however it was authored -- most
+ * often a layered ASS event stacking a shadow copy under the visible one.
+ */
+function collapseExactDuplicates(cues: AnnotatedSubtitleCue[]): AnnotatedSubtitleCue[] {
+  const seen = new Set<string>();
+  return cues.filter((cue) => {
+    const key = cueKey(cue);
+    if (seen.has(key)) {
+      return false;
+    }
+    seen.add(key);
+    return true;
+  });
+}
+
+function countFramesShorterThan(run: AnnotatedSubtitleCue[], maxSeconds: number): number {
+  return run.filter((cue) => cue.endTime - cue.startTime < maxSeconds).length;
+}
+
+/**
+ * Evidence that a run of ASS events is one animation rather than several authored lines.
+ * A static tag says nothing on its own -- three events sharing one `\clip(...)` are three
+ * signs -- so the tag has to be temporal by nature (`\t`, `\move`, karaoke timing, or
+ * anything wrapped in `\t(...)`), an animated `Effect` column, or a value that actually
+ * changes from event to event, which is how per-frame typesetting is authored.
+ */
+export function hasAssAnimationEvidence(run: AnnotatedSubtitleCue[]): boolean {
+  if (run.every((cue) => hasAssTemporalOverride(cue.overrides))) {
+    return true;
+  }
+  if (run.every((cue) => isAnimatedAssEffectKind(cue.effectKind))) {
+    return true;
+  }
+
+  const [first] = run;
+  const everyEventTypeset = run.every((cue) => cue.overrides.length > 0);
+  const signatureChanges = run.some((cue) => cue.overrideSignature !== first!.overrideSignature);
+  return everyEventTypeset && signatureChanges;
+}
+
+export function isAnimationBurst(
+  run: AnnotatedSubtitleCue[],
+  format: SubtitleSourceFormat,
+): boolean {
+  if (run.length < MIN_BURST_EVENTS) {
+    return false;
+  }
+
+  if (format === 'srt') {
+    return (
+      run.length >= MIN_TIMING_ONLY_FRAMES &&
+      countFramesShorterThan(run, TIMING_ONLY_FRAME_MAX_SECONDS) === run.length
+    );
+  }
+
+  if (countFramesShorterThan(run, ANIMATION_FRAME_MAX_SECONDS) < MIN_TAGGED_BURST_FRAMES) {
+    return false;
+  }
+
+  // One animation belongs to one styled, one named source line. Two characters trading
+  // the same short word are two styles or two actors, and never merge.
+  const [first] = run;
+  if (run.some((cue) => cue.style !== first!.style || cue.name !== first!.name)) {
+    return false;
+  }
+
+  return hasAssAnimationEvidence(run);
+}
+
+/**
+ * Karaoke and sign typesetting emits one Dialogue event per animation frame, all carrying
+ * the same visible text over a contiguous span. Collapse each such run into a single cue.
+ *
+ * Only runs that look like animation collapse. Two ordinary lines that happen to repeat
+ * -- several characters each saying `おはよう` in turn, a positioned sign redrawn with a
+ * different fade -- stay separate, because merging them would destroy real mineable lines.
+ */
+function collapseAnimationBursts(
+  cues: AnnotatedSubtitleCue[],
+  format: SubtitleSourceFormat,
+): AnnotatedSubtitleCue[] {
+  const indicesByText = new Map<string, number[]>();
+  cues.forEach((cue, index) => {
+    const bucket = indicesByText.get(cue.text);
+    if (bucket) {
+      bucket.push(index);
+    } else {
+      indicesByText.set(cue.text, [index]);
+    }
+  });
+
+  const dropped = new Set<number>();
+  const extendedEnd = new Map<number, number>();
+
+  for (const indices of indicesByText.values()) {
+    if (indices.length < MIN_BURST_EVENTS) {
+      continue;
+    }
+
+    let runStart = 0;
+    while (runStart < indices.length) {
+      let runEnd = runStart;
+      let chainEnd = cues[indices[runStart]!]!.endTime;
+
+      while (runEnd + 1 < indices.length) {
+        const next = cues[indices[runEnd + 1]!]!;
+        if (next.startTime > chainEnd + DUPLICATE_CUE_GAP_TOLERANCE_SECONDS) {
+          break;
+        }
+        chainEnd = Math.max(chainEnd, next.endTime);
+        runEnd += 1;
+      }
+
+      const run = indices.slice(runStart, runEnd + 1).map((index) => cues[index]!);
+      if (isAnimationBurst(run, format)) {
+        for (let i = runStart + 1; i <= runEnd; i += 1) {
+          dropped.add(indices[i]!);
+        }
+        extendedEnd.set(indices[runStart]!, chainEnd);
+      }
+
+      runStart = runEnd + 1;
+    }
+  }
+
+  if (dropped.size === 0) {
+    return cues;
+  }
+
+  const merged: AnnotatedSubtitleCue[] = [];
+  cues.forEach((cue, index) => {
+    if (dropped.has(index)) {
+      return;
+    }
+    const end = extendedEnd.get(index);
+    merged.push(end !== undefined && end > cue.endTime ? { ...cue, endTime: end } : cue);
+  });
+  return merged;
+}
+
+/**
+ * Collapse redundant cues. Input must already be sorted by non-decreasing `startTime`,
+ * ties broken by `endTime` then source `order` -- burst detection chains events by
+ * comparing each one against the running end of the events before it, so an unsorted
+ * list breaks runs apart and leaves the frames behind.
+ */
+export function mergeDuplicateCues(
+  cues: AnnotatedSubtitleCue[],
+  format: SubtitleSourceFormat,
+): AnnotatedSubtitleCue[] {
+  return collapseAnimationBursts(collapseExactDuplicates(cues), format);
+}

src/core/services/subtitle-cue-parser.test.ts

@@ -91,6 +91,17 @@ test('parseSrtCues skips malformed timing lines gracefully', () => {
   assert.equal(cues[0]!.text, '有効');
 });
 
+test('parseSubtitleCues strips complete brace blocks from SRT and VTT text', () => {
+  const content = ['1', '00:00:01,000 --> 00:00:02,000', '彼は{謎}と言った', ''].join('\n');
+
+  for (const filename of ['test.srt', 'test.vtt']) {
+    const cues = parseSubtitleCues(content, filename);
+
+    assert.equal(cues.length, 1, filename);
+    assert.equal(cues[0]!.text, '彼はと言った', filename);
+  }
+});
+
 test('parseAssCues parses basic ASS dialogue lines', () => {
   const content = [
     '[Script Info]',
@@ -137,7 +148,9 @@ test('parseAssCues handles text containing commas', () => {
   assert.equal(cues[0]!.text, 'はい、そうです、ね');
 });
 
-test('parseAssCues handles \\N line breaks', () => {
+test('parseAssCues decodes \\N line breaks into real newlines', () => {
+  // ASS is decoded once, here at ingestion, so cue text matches what mpv hands over for
+  // the same line played live.
   const content = [
     '[Events]',
     'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
@@ -146,7 +159,7 @@ test('parseAssCues handles \\N line breaks', () => {
 
   const cues = parseAssCues(content);
 
-  assert.equal(cues[0]!.text, '一行目\\N二行目');
+  assert.equal(cues[0]!.text, '一行目\n二行目');
 });
 
 test('parseAssCues strips HTML-like markup while preserving ASS line breaks', () => {
@@ -158,7 +171,46 @@ test('parseAssCues strips HTML-like markup while preserving ASS line breaks', ()
 
   const cues = parseAssCues(content);
 
-  assert.equal(cues[0]!.text, '一行目\\N二行目');
+  assert.equal(cues[0]!.text, '一行目\n二行目');
+});
+
+test('parseAssCues drops vector drawing runs enabled by \\p', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 1,0:00:01.00,0:00:04.00,Default,,0,0,0,,{\\an5\\pos(730,1042)\\p1\\blur1}m 20 0 b 10 0 0 10 0 20 b 0 31 10 40 20 40 {\\p0}',
+    'Dialogue: 0,0:00:05.00,0:00:08.00,Default,,0,0,0,,これは字幕',
+  ].join('\n');
+
+  const cues = parseAssCues(content);
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.text, 'これは字幕');
+});
+
+test('parseAssCues keeps text that follows a \\p0 reset on the same line', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:04.00,Default,,0,0,0,,{\\p1}m 0 0 l 10 10{\\p0}本文{\\p1}m 5 5 l 6 6{\\p0}続き',
+  ].join('\n');
+
+  const cues = parseAssCues(content);
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.text, '本文続き');
+});
+
+test('parseAssCues leaves \\pos untouched when no drawing mode is active', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:04.00,Default,,0,0,0,,{\\pos(960,1068)\\bord3}位置指定',
+  ].join('\n');
+
+  const cues = parseAssCues(content);
+
+  assert.equal(cues[0]!.text, '位置指定');
 });
 
 test('parseAssCues returns empty for content without Events section', () => {
@@ -258,6 +310,344 @@ test('parseSubtitleCues returns cues sorted by start time', () => {
   assert.equal(cues[1]!.text, '二番目');
 });
 
+test('parseSubtitleCues collapses per-frame karaoke duplicates into one cue', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.05,OP_JP,,0,0,0,,{\\clip(m 1 1)}過ぎ去ってしまう瞬間を',
+    'Dialogue: 0,0:00:01.05,0:00:01.09,OP_JP,,0,0,0,,{\\clip(m 2 2)}過ぎ去ってしまう瞬間を',
+    'Dialogue: 0,0:00:01.09,0:00:03.55,OP_JP,,0,0,0,,{\\clip(m 3 3)}過ぎ去ってしまう瞬間を',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.startTime, 1.0);
+  assert.equal(cues[0]!.endTime, 3.55);
+  assert.equal(cues[0]!.text, '過ぎ去ってしまう瞬間を');
+});
+
+test('parseSubtitleCues keeps back-to-back plain dialogue repeats separate', () => {
+  // Several characters greeting in turn: distinct utterances that happen to abut.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:04:05.67,0:04:06.82,Dial_JP,,0,0,0,,おはよう',
+    'Dialogue: 0,0:04:06.82,0:04:07.56,Dial_JP,,0,0,0,,おはよう',
+    'Dialogue: 0,0:04:07.56,0:04:08.78,Dial_JP,,0,0,0,,おはよう',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+  assert.equal(cues[0]!.endTime, 246.82);
+  assert.equal(cues[2]!.startTime, 247.56);
+});
+
+test('parseSubtitleCues collapses exact duplicate cues even without effect tags', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:04.00,Default,,0,0,0,,重なった行',
+    'Dialogue: 1,0:00:01.00,0:00:04.00,Default,,0,0,0,,重なった行',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 1);
+});
+
+test('parseSubtitleCues collapses tag-less animation frames in converted SRT', () => {
+  // ASS -> SRT conversion drops override tags, so only the ~0.04s frame timing remains.
+  const lines = ['1', '00:00:07,870 --> 00:00:07,910', 'Kaguya Wants to be Confessed to', ''];
+  for (let i = 1; i < 8; i++) {
+    const start = 7910 + (i - 1) * 40;
+    const end = start + 40;
+    const at = (ms: number) =>
+      `00:00:0${Math.floor(ms / 1000)},${String(ms % 1000).padStart(3, '0')}`;
+    lines.push(String(i + 1), `${at(start)} --> ${at(end)}`, 'Kaguya Wants to be Confessed to', '');
+  }
+
+  const cues = parseSubtitleCues(lines.join('\n'), 'test.srt');
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.startTime, 7.87);
+});
+
+test('parseSubtitleCues keeps identical lines that recur far apart', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:02.00,Default,,0,0,0,,なんで',
+    'Dialogue: 0,0:05:00.00,0:05:01.00,Default,,0,0,0,,なんで',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 2);
+  assert.equal(cues[0]!.startTime, 1.0);
+  assert.equal(cues[1]!.startTime, 300.0);
+});
+
+test('parseSubtitleCues keeps two positioned signs that repeat the same text', () => {
+  // Both carry override tags, but `\pos` and `\fad` are static placement, not animation.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:01:00.00,0:01:03.00,Sign,,0,0,0,,{\\pos(960,120)\\fad(200,200)}第一話',
+    'Dialogue: 0,0:01:03.00,0:01:06.00,Sign,,0,0,0,,{\\pos(960,900)\\fad(200,200)}第一話',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 2);
+  assert.equal(cues[1]!.startTime, 63.0);
+});
+
+test('parseSubtitleCues keeps a run of ordinary positioned lines separate', () => {
+  // Three events is a sequence, but none of them runs at animation-frame speed.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:01:00.00,0:01:02.00,Sign,,0,0,0,,{\\pos(960,120)\\fad(100,100)}止まれ',
+    'Dialogue: 0,0:01:02.00,0:01:04.00,Sign,,0,0,0,,{\\pos(960,120)\\fad(100,100)}止まれ',
+    'Dialogue: 0,0:01:04.00,0:01:06.00,Sign,,0,0,0,,{\\pos(960,120)\\fad(100,100)}止まれ',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+});
+
+test('parseSubtitleCues keeps a short repeated SRT pair without burst evidence', () => {
+  const content = [
+    '1',
+    '00:00:01,000 --> 00:00:01,200',
+    'えっ',
+    '',
+    '2',
+    '00:00:01,200 --> 00:00:01,400',
+    'えっ',
+    '',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.srt');
+
+  assert.equal(cues.length, 2);
+});
+
+test('parseSubtitleCues collapses a burst marked only by the Effect column', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.05,OP_JP,,0,0,0,Karaoke,歌詞',
+    'Dialogue: 0,0:00:01.05,0:00:01.09,OP_JP,,0,0,0,Karaoke,歌詞',
+    'Dialogue: 0,0:00:01.09,0:00:03.55,OP_JP,,0,0,0,Karaoke,歌詞',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.endTime, 3.55);
+});
+
+test('parseSubtitleCues keeps a second karaoke burst that starts after a gap', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.05,OP_JP,,0,0,0,,{\\clip(m 1 1)}リフレイン',
+    'Dialogue: 0,0:00:01.05,0:00:01.09,OP_JP,,0,0,0,,{\\clip(m 2 2)}リフレイン',
+    'Dialogue: 0,0:00:01.09,0:00:03.00,OP_JP,,0,0,0,,{\\clip(m 3 3)}リフレイン',
+    'Dialogue: 0,0:00:20.00,0:00:20.05,OP_JP,,0,0,0,,{\\clip(m 1 1)}リフレイン',
+    'Dialogue: 0,0:00:20.05,0:00:20.09,OP_JP,,0,0,0,,{\\clip(m 2 2)}リフレイン',
+    'Dialogue: 0,0:00:20.09,0:00:22.00,OP_JP,,0,0,0,,{\\clip(m 3 3)}リフレイン',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 2);
+  assert.equal(cues[0]!.endTime, 3.0);
+  assert.equal(cues[1]!.startTime, 20.0);
+  assert.equal(cues[1]!.endTime, 22.0);
+});
+
+test('parseSubtitleCues does not merge a burst into unrelated dialogue between frames', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.05,OP_JP,,0,0,0,,{\\clip(m 1 1)}歌詞',
+    'Dialogue: 0,0:00:01.02,0:00:03.00,Dial_JP,,0,0,0,,別のセリフ',
+    'Dialogue: 0,0:00:01.05,0:00:01.09,OP_JP,,0,0,0,,{\\clip(m 2 2)}歌詞',
+    'Dialogue: 0,0:00:01.09,0:00:03.55,OP_JP,,0,0,0,,{\\clip(m 3 3)}歌詞',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 2);
+  assert.deepEqual(
+    cues.map((cue) => cue.text),
+    ['歌詞', '別のセリフ'],
+  );
+  assert.equal(cues[0]!.endTime, 3.55);
+});
+
+test('parseSubtitleCues keeps rapid ASS lines from different actors separate', () => {
+  // Three 200ms `えっ` reactions traded between characters. Fast, adjacent and identical,
+  // but authored as three lines: different styles and different actors.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Dial_A,アリス,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Dial_B,ボブ,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.40,0:00:01.60,Dial_C,キャロル,0,0,0,,えっ',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+});
+
+test('parseSubtitleCues reads the speaker column when it is spelled Actor', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Actor, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Dial_JP,アリス,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Dial_JP,ボブ,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.40,0:00:01.60,Dial_JP,キャロル,0,0,0,,えっ',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+});
+
+test('parseSubtitleCues does not treat a custom Effect name as animation', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Sign,,0,0,0,scrolling-credit,制作',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Sign,,0,0,0,scrolling-credit,制作',
+    'Dialogue: 0,0:00:01.40,0:00:01.60,Sign,,0,0,0,scrolling-credit,制作',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+});
+
+test('parseSubtitleCues keeps rapid ASS lines that share a style but not an actor', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Dial_JP,アリス,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Dial_JP,ボブ,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.40,0:00:01.60,Dial_JP,キャロル,0,0,0,,えっ',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+});
+
+test('parseSubtitleCues keeps untagged rapid ASS repeats separate', () => {
+  // No overrides at all: timing-only evidence is an SRT/VTT fallback and must not apply
+  // to ASS, where the absence of typesetting is itself evidence of plain dialogue.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.05,Dial_JP,,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.05,0:00:01.10,Dial_JP,,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.10,0:00:01.15,Dial_JP,,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.15,0:00:01.20,Dial_JP,,0,0,0,,えっ',
+    'Dialogue: 0,0:00:01.20,0:00:01.25,Dial_JP,,0,0,0,,えっ',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 5);
+});
+
+test('parseSubtitleCues keeps repeated signs sharing one static clip', () => {
+  // `\clip` is a static shape for the event. Three events with the identical clip were
+  // typeset the same way, so none of them is a frame of the others.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Sign,,0,0,0,,{\\clip(0,0,100,100)}注意',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Sign,,0,0,0,,{\\clip(0,0,100,100)}注意',
+    'Dialogue: 0,0:00:01.40,0:00:01.60,Sign,,0,0,0,,{\\clip(0,0,100,100)}注意',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 3);
+});
+
+test('parseSubtitleCues collapses a sign animated through \\t', () => {
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Sign,,0,0,0,,{\\pos(10,10)\\t(0,200,\\frz30)}回る',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Sign,,0,0,0,,{\\pos(10,10)\\t(0,200,\\frz30)}回る',
+    'Dialogue: 0,0:00:01.40,0:00:03.00,Sign,,0,0,0,,{\\pos(10,10)\\t(0,200,\\frz30)}回る',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.ass');
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.endTime, 3.0);
+});
+
+test('parseSubtitleCues keeps a short repeated SRT run above the frame threshold', () => {
+  // Five contiguous 200ms cues: a sequence, but nowhere near animation-frame speed.
+  const lines: string[] = [];
+  for (let i = 0; i < 5; i++) {
+    const start = 1000 + i * 200;
+    const at = (ms: number) =>
+      `00:00:0${Math.floor(ms / 1000)},${String(ms % 1000).padStart(3, '0')}`;
+    lines.push(String(i + 1), `${at(start)} --> ${at(start + 200)}`, 'えっ', '');
+  }
+
+  const cues = parseSubtitleCues(lines.join('\n'), 'test.srt');
+
+  assert.equal(cues.length, 5);
+});
+
+test('parseSubtitleCues keeps a short SRT frame run below the minimum length', () => {
+  // Four 40ms frames: frame-speed, but too few to tell an animation from an artefact.
+  const lines: string[] = [];
+  for (let i = 0; i < 4; i++) {
+    const start = 7870 + i * 40;
+    const at = (ms: number) =>
+      `00:00:0${Math.floor(ms / 1000)},${String(ms % 1000).padStart(3, '0')}`;
+    lines.push(String(i + 1), `${at(start)} --> ${at(start + 40)}`, 'タイトル', '');
+  }
+
+  const cues = parseSubtitleCues(lines.join('\n'), 'test.srt');
+
+  assert.equal(cues.length, 4);
+});
+
+test('parseSubtitleCues applies ASS burst rules to ASS content behind an .srt filename', () => {
+  // The extension lies, so the SRT parser finds nothing and the content-sniffing fallback
+  // takes over -- which has to carry the `ass` source format with it, or the far stricter
+  // timing-only thresholds would let this karaoke burst through as three cues.
+  const content = [
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+    'Dialogue: 0,0:00:01.00,0:00:01.20,Karaoke,,0,0,0,,{\\k20}歌詞',
+    'Dialogue: 0,0:00:01.20,0:00:01.40,Karaoke,,0,0,0,,{\\k20}歌詞',
+    'Dialogue: 0,0:00:01.40,0:00:03.00,Karaoke,,0,0,0,,{\\k20}歌詞',
+  ].join('\n');
+
+  const cues = parseSubtitleCues(content, 'test.srt');
+
+  assert.equal(cues.length, 1);
+  assert.equal(cues[0]!.startTime, 1.0);
+  assert.equal(cues[0]!.endTime, 3.0);
+  assert.equal(cues[0]!.text, '歌詞');
+});
+
 test('parseSubtitleCues detects subtitle formats from remote URLs', () => {
   const assContent = [
     '[Events]',

src/core/services/subtitle-cue-parser.ts

@@ -1,9 +1,46 @@
+import {
+  assOverrideSignature,
+  assToPlainText,
+  collectAssOverrideCommands,
+  parseAssEffectField,
+  type AssEffectKind,
+  type AssOverrideCommand,
+} from './ass-text';
+import { mergeDuplicateCues } from './subtitle-cue-dedup';
+
 export interface SubtitleCue {
   startTime: number;
   endTime: number;
   text: string;
 }
 
+/**
+ * Everything the parser knows about a source event, shared only with the dedup engine.
+ * Deduplication needs the authoring context -- which style the line belongs to, which
+ * override commands it carries, whether the `Effect` column was set -- to tell a karaoke
+ * burst apart from two characters saying the same word in turn. None of it is meaningful
+ * outside the parser, so the public API stays `{startTime, endTime, text}`.
+ */
+export interface AnnotatedSubtitleCue extends SubtitleCue {
+  /** Text exactly as authored, override blocks and all. */
+  rawText: string;
+  style: string;
+  layer: number;
+  /** ASS `Name`/`Actor` column. */
+  name: string;
+  /** ASS `Effect` column, verbatim. */
+  effect: string;
+  effectKind: AssEffectKind;
+  /** Override commands found in `{...}` blocks, with their arguments. */
+  overrides: readonly AssOverrideCommand[];
+  /** Canonical form of `overrides`, for spotting values that change across a run. */
+  overrideSignature: string;
+  /** Position in the source file, so sorting by time stays deterministic across layers. */
+  order: number;
+}
+
+export type SubtitleSourceFormat = 'ass' | 'srt';
+
 const HTML_SUBTITLE_TAG_PATTERN = /<\/?[A-Za-z][^>\n]*>/g;
 
 const SRT_TIMING_PATTERN =
@@ -23,12 +60,21 @@ function parseTimestamp(
   );
 }
 
+/**
+ * The single ASS decode for the file path: cues leave the parser as plain text with real
+ * line breaks, matching what mpv hands over for the same line played live. No layer
+ * downstream decodes ASS again.
+ */
 function sanitizeSubtitleCueText(text: string): string {
-  return text.replace(ASS_OVERRIDE_TAG_PATTERN, '').replace(HTML_SUBTITLE_TAG_PATTERN, '').trim();
+  return assToPlainText(text, '\n').replace(HTML_SUBTITLE_TAG_PATTERN, '').trim();
 }
 
-export function parseSrtCues(content: string): SubtitleCue[] {
+function toPublicCues(cues: AnnotatedSubtitleCue[]): SubtitleCue[] {
-  const cues: SubtitleCue[] = [];
+  return cues.map(({ startTime, endTime, text }) => ({ startTime, endTime, text }));
+}
+
+function parseAnnotatedSrtCues(content: string): AnnotatedSubtitleCue[] {
+  const cues: AnnotatedSubtitleCue[] = [];
   const lines = content.split(/\r?\n/);
   let i = 0;
 
@@ -60,20 +106,39 @@ export function parseSrtCues(content: string): SubtitleCue[] {
       i += 1;
     }
 
-    const text = sanitizeSubtitleCueText(textLines.join('\n'));
+    const rawText = textLines.join('\n');
+    const text = sanitizeSubtitleCueText(rawText);
     if (text) {
-      cues.push({ startTime, endTime, text });
+      cues.push({
+        startTime,
+        endTime,
+        text,
+        rawText,
+        style: '',
+        layer: 0,
+        name: '',
+        effect: '',
+        effectKind: 'none',
+        // SRT and VTT carry no authoring metadata, and the dedup engine never reads
+        // overrides for those formats -- collecting them would be parsing for nobody.
+        overrides: [],
+        overrideSignature: '',
+        order: cues.length,
+      });
     }
   }
 
   return cues;
 }
 
-const ASS_OVERRIDE_TAG_PATTERN = /\{[^}]*\}/g;
+export function parseSrtCues(content: string): SubtitleCue[] {
+  return toPublicCues(parseAnnotatedSrtCues(content));
+}
 
 const ASS_TIMING_PATTERN = /^(\d+):(\d{2}):(\d{2})\.(\d{1,2})$/;
 const ASS_FORMAT_PREFIX = 'Format:';
 const ASS_DIALOGUE_PREFIX = 'Dialogue:';
+const ASS_NAME_FIELD_ALIASES = ['name', 'actor'];
 
 function parseAssTimestamp(raw: string): number | null {
   const match = ASS_TIMING_PATTERN.exec(raw.trim());
@@ -87,13 +152,43 @@ function parseAssTimestamp(raw: string): number | null {
   return hours * 3600 + minutes * 60 + seconds + centiseconds / 100;
 }
 
-export function parseAssCues(content: string): SubtitleCue[] {
+function readField(fields: string[], index: number): string {
-  const cues: SubtitleCue[] = [];
+  return index >= 0 && index < fields.length ? fields[index]!.trim() : '';
+}
+
+function findFieldIndex(formatFields: string[], aliases: string[]): number {
+  for (const alias of aliases) {
+    const index = formatFields.indexOf(alias);
+    if (index >= 0) {
+      return index;
+    }
+  }
+  return -1;
+}
+
+function parseAnnotatedAssCues(content: string): AnnotatedSubtitleCue[] {
+  const cues: AnnotatedSubtitleCue[] = [];
   const lines = content.split(/\r?\n/);
   let inEventsSection = false;
-  let startFieldIndex = -1;
+  const fieldIndex = {
-  let endFieldIndex = -1;
+    start: -1,
-  let textFieldIndex = -1;
+    end: -1,
+    text: -1,
+    style: -1,
+    layer: -1,
+    name: -1,
+    effect: -1,
+  };
+
+  const resetFieldIndex = () => {
+    fieldIndex.start = -1;
+    fieldIndex.end = -1;
+    fieldIndex.text = -1;
+    fieldIndex.style = -1;
+    fieldIndex.layer = -1;
+    fieldIndex.name = -1;
+    fieldIndex.effect = -1;
+  };
 
   for (const line of lines) {
     const trimmed = line.trim();
@@ -101,9 +196,7 @@ export function parseAssCues(content: string): SubtitleCue[] {
     if (trimmed.startsWith('[') && trimmed.endsWith(']')) {
       inEventsSection = trimmed.toLowerCase() === '[events]';
       if (!inEventsSection) {
-        startFieldIndex = -1;
+        resetFieldIndex();
-        endFieldIndex = -1;
-        textFieldIndex = -1;
       }
       continue;
     }
@@ -117,9 +210,15 @@ export function parseAssCues(content: string): SubtitleCue[] {
         .slice(ASS_FORMAT_PREFIX.length)
         .split(',')
         .map((field) => field.trim().toLowerCase());
-      startFieldIndex = formatFields.indexOf('start');
+      fieldIndex.start = formatFields.indexOf('start');
-      endFieldIndex = formatFields.indexOf('end');
+      fieldIndex.end = formatFields.indexOf('end');
-      textFieldIndex = formatFields.indexOf('text');
+      fieldIndex.text = formatFields.indexOf('text');
+      fieldIndex.style = formatFields.indexOf('style');
+      fieldIndex.layer = formatFields.indexOf('layer');
+      // Aegisub writes the speaker column as `Actor`; the v4+ spec calls it `Name`.
+      // Missing it costs the burst check its speaker guard, so both spellings count.
+      fieldIndex.name = findFieldIndex(formatFields, ASS_NAME_FIELD_ALIASES);
+      fieldIndex.effect = formatFields.indexOf('effect');
       continue;
     }
 
@@ -127,34 +226,57 @@ export function parseAssCues(content: string): SubtitleCue[] {
       continue;
     }
 
-    if (startFieldIndex < 0 || endFieldIndex < 0 || textFieldIndex < 0) {
+    if (fieldIndex.start < 0 || fieldIndex.end < 0 || fieldIndex.text < 0) {
       continue;
     }
 
     const fields = trimmed.slice(ASS_DIALOGUE_PREFIX.length).split(',');
     if (
-      startFieldIndex >= fields.length ||
+      fieldIndex.start >= fields.length ||
-      endFieldIndex >= fields.length ||
+      fieldIndex.end >= fields.length ||
-      textFieldIndex >= fields.length
+      fieldIndex.text >= fields.length
     ) {
       continue;
     }
 
-    const startTime = parseAssTimestamp(fields[startFieldIndex]!);
+    const startTime = parseAssTimestamp(fields[fieldIndex.start]!);
-    const endTime = parseAssTimestamp(fields[endFieldIndex]!);
+    const endTime = parseAssTimestamp(fields[fieldIndex.end]!);
     if (startTime === null || endTime === null) {
       continue;
     }
 
-    const text = sanitizeSubtitleCueText(fields.slice(textFieldIndex).join(','));
+    const rawText = fields.slice(fieldIndex.text).join(',');
-    if (text) {
+    const text = sanitizeSubtitleCueText(rawText);
-      cues.push({ startTime, endTime, text });
+    if (!text) {
+      continue;
     }
+
+    const effect = readField(fields, fieldIndex.effect);
+    const layer = Number(readField(fields, fieldIndex.layer));
+    const overrides = collectAssOverrideCommands(rawText);
+    cues.push({
+      startTime,
+      endTime,
+      text,
+      rawText,
+      style: readField(fields, fieldIndex.style),
+      layer: Number.isFinite(layer) ? layer : 0,
+      name: readField(fields, fieldIndex.name),
+      effect,
+      effectKind: parseAssEffectField(effect),
+      overrides,
+      overrideSignature: assOverrideSignature(overrides),
+      order: cues.length,
+    });
   }
 
   return cues;
 }
 
+export function parseAssCues(content: string): SubtitleCue[] {
+  return toPublicCues(parseAnnotatedAssCues(content));
+}
+
 function detectSubtitleFormat(source: string): 'srt' | 'vtt' | 'ass' | 'ssa' | null {
   const [normalizedSource = source] =
     (() => {
@@ -173,27 +295,31 @@ function detectSubtitleFormat(source: string): 'srt' | 'vtt' | 'ass' | 'ssa' | n
 
 export function parseSubtitleCues(content: string, filename: string): SubtitleCue[] {
   const format = detectSubtitleFormat(filename);
-  let cues: SubtitleCue[];
+  let cues: AnnotatedSubtitleCue[];
+  let sourceFormat: SubtitleSourceFormat = 'srt';
 
   switch (format) {
     case 'srt':
     case 'vtt':
-      cues = parseSrtCues(content);
+      cues = parseAnnotatedSrtCues(content);
       break;
     case 'ass':
     case 'ssa':
-      cues = parseAssCues(content);
+      cues = parseAnnotatedAssCues(content);
+      sourceFormat = 'ass';
       break;
     default:
       cues = [];
   }
 
   if (cues.length === 0) {
-    const assCues = parseAssCues(content);
+    const assCues = parseAnnotatedAssCues(content);
-    const srtCues = parseSrtCues(content);
+    const srtCues = parseAnnotatedSrtCues(content);
-    cues = assCues.length >= srtCues.length ? assCues : srtCues;
+    const preferAss = assCues.length >= srtCues.length;
+    cues = preferAss ? assCues : srtCues;
+    sourceFormat = preferAss && assCues.length > 0 ? 'ass' : 'srt';
   }
 
-  cues.sort((a, b) => a.startTime - b.startTime);
+  cues.sort((a, b) => a.startTime - b.startTime || a.endTime - b.endTime || a.order - b.order);
-  return cues;
+  return toPublicCues(mergeDuplicateCues(cues, sourceFormat));
 }

src/core/services/subtitle-prefetch.test.ts

@@ -74,7 +74,6 @@ test('prefetch service tokenizes priority window cues and caches them', async ()
     preCacheTokenization: (text, data) => {
       cached.set(text, data);
     },
-    isCacheFull: () => false,
     priorityWindowSize: 3,
   });
 
@@ -91,32 +90,38 @@ test('prefetch service tokenizes priority window cues and caches them', async ()
   assert.ok(cached.has('line-2'));
 });
 
-test('prefetch service stops when cache is full', async () => {
+test('prefetch service warms every cue even when the cache evicts along the way', async () => {
   const cues = makeCues(20);
-  let tokenizeCalls = 0;
+  const tokenized: string[] = [];
-  let cacheSize = 0;
+  // Stand-in for the LRU: only the last 5 entries survive, so later cues evict earlier ones.
+  const cache = new Set<string>();
 
   const service = createSubtitlePrefetchService({
     cues,
     tokenizeSubtitle: async (text) => {
-      tokenizeCalls += 1;
+      tokenized.push(text);
       return { text, tokens: [] };
     },
-    preCacheTokenization: () => {
+    preCacheTokenization: (text) => {
-      cacheSize += 1;
+      cache.add(text);
+      while (cache.size > 5) {
+        const oldest = cache.values().next().value;
+        if (oldest === undefined) break;
+        cache.delete(oldest);
+      }
     },
-    isCacheFull: () => cacheSize >= 5,
+    hasCachedTokenization: (text) => cache.has(text),
     priorityWindowSize: 3,
   });
 
   service.start(0);
-  for (let i = 0; i < 30; i += 1) {
+  for (let i = 0; i < 60; i += 1) {
     await flushMicrotasks();
   }
   service.stop();
 
-  // Should have stopped at 5 (cache full), not tokenized all 20
+  assert.equal(tokenized.length, 20, `Expected all 20 cues warmed, got ${tokenized.length}`);
-  assert.ok(tokenizeCalls <= 6, `Expected <= 6 tokenize calls, got ${tokenizeCalls}`);
+  assert.equal(new Set(tokenized).size, 20, 'Each cue is tokenized at most once per run');
 });
 
 test('prefetch service can be stopped mid-flight', async () => {
@@ -130,7 +135,6 @@ test('prefetch service can be stopped mid-flight', async () => {
       return { text, tokens: [] };
     },
     preCacheTokenization: () => {},
-    isCacheFull: () => false,
     priorityWindowSize: 3,
   });
 
@@ -159,7 +163,6 @@ test('prefetch service onSeek re-prioritizes from new position', async () => {
     preCacheTokenization: (text) => {
       cachedTexts.push(text);
     },
-    isCacheFull: () => false,
     priorityWindowSize: 3,
   });
 
@@ -183,7 +186,7 @@ test('prefetch service onSeek re-prioritizes from new position', async () => {
   assert.ok(hasPostSeekCue, 'Should have cached cues after seek position');
 });
 
-test('prefetch service still warms the priority window when cache is full', async () => {
+test('prefetch service warms the priority window ahead of the rest of the file', async () => {
   const cues = makeCues(20);
   const cachedTexts: string[] = [];
 
@@ -193,7 +196,6 @@ test('prefetch service still warms the priority window when cache is full', asyn
     preCacheTokenization: (text) => {
       cachedTexts.push(text);
     },
-    isCacheFull: () => true,
     priorityWindowSize: 3,
   });
 
@@ -217,7 +219,6 @@ test('prefetch service pause/resume halts and continues tokenization', async ()
       return { text, tokens: [] };
     },
     preCacheTokenization: () => {},
-    isCacheFull: () => false,
     priorityWindowSize: 3,
   });
 
@@ -255,7 +256,6 @@ test('prefetch service skips cues already present in tokenization cache', async
     },
     preCacheTokenization: () => {},
     hasCachedTokenization: (text) => text === 'line-0' || text === 'line-1',
-    isCacheFull: () => false,
     priorityWindowSize: 3,
   });
 
@@ -285,7 +285,6 @@ test('prefetch service deduplicates repeated cue text within a run', async () =>
       return { text, tokens: [] };
     },
     preCacheTokenization: () => {},
-    isCacheFull: () => false,
     priorityWindowSize: 3,
   });
 

src/core/services/subtitle-prefetch.ts

@@ -7,7 +7,6 @@ export interface SubtitlePrefetchServiceDeps {
   tokenizeSubtitle: (text: string) => Promise<SubtitleData | null>;
   preCacheTokenization: (text: string, data: SubtitleData) => void;
   hasCachedTokenization?: (text: string) => boolean;
-  isCacheFull: () => boolean;
   priorityWindowSize?: number;
 }
 
@@ -57,11 +56,14 @@ export function createSubtitlePrefetchService(
   let paused = false;
   let currentRunId = 0;
 
+  // A run is a single bounded pass over one file's cues, deduped by `warmedKeys` and by
+  // `hasCachedTokenization`, so the worst case is one tokenization per cue. The cache is
+  // an LRU and bounds its own memory, so a full cache is not a reason to stop warming;
+  // stopping there used to leave the tail of longer media permanently uncached.
   async function tokenizeCueList(
     cuesToProcess: SubtitleCue[],
     runId: number,
     warmedKeys: Set<string>,
-    options: { allowWhenCacheFull?: boolean } = {},
   ): Promise<void> {
     for (const cue of cuesToProcess) {
       if (stopped || runId !== currentRunId) {
@@ -77,10 +79,6 @@ export function createSubtitlePrefetchService(
         return;
       }
 
-      if (!options.allowWhenCacheFull && deps.isCacheFull()) {
-        return;
-      }
-
       const cacheKey = normalizeSubtitleCacheKey(cue.text);
       if (!cacheKey || warmedKeys.has(cacheKey) || deps.hasCachedTokenization?.(cue.text)) {
         if (cacheKey) {
@@ -110,7 +108,7 @@ export function createSubtitlePrefetchService(
 
     // Phase 1: Priority window
     const priorityCues = computePriorityWindow(cues, currentTimeSeconds, windowSize);
-    await tokenizeCueList(priorityCues, runId, warmedKeys, { allowWhenCacheFull: true });
+    await tokenizeCueList(priorityCues, runId, warmedKeys);
 
     if (stopped || runId !== currentRunId) {
       return;

src/core/services/subtitle-processing-controller-race.test.ts

@@ -0,0 +1,159 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import type { SubtitleData } from '../../types';
+import { createSubtitleProcessingController } from './subtitle-processing-controller';
+
+function flushMicrotasks(): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, 0));
+}
+
+test('new subtitle emits plain immediately without parallel tokenization or a stale overwrite', async () => {
+  const emitted: SubtitleData[] = [];
+  const resolvers = new Map<string, (value: SubtitleData | null) => void>();
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) =>
+      await new Promise<SubtitleData | null>((resolve) => {
+        resolvers.set(text, resolve);
+      }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('first');
+  controller.onSubtitleChange('second');
+
+  assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
+    { text: 'second', tokens: null },
+  ]);
+  assert.equal(resolvers.has('second'), false);
+
+  const resolveFirst = resolvers.get('first');
+  assert.ok(resolveFirst);
+  resolveFirst({ text: 'first', tokens: [] });
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
+    { text: 'second', tokens: null },
+  ]);
+  assert.equal(resolvers.has('second'), true);
+
+  const resolveSecond = resolvers.get('second');
+  assert.ok(resolveSecond);
+  resolveSecond({ text: 'second', tokens: [] });
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
+    { text: 'second', tokens: null },
+    { text: 'second', tokens: [] },
+  ]);
+});
+
+test('subtitle clears immediately while previous tokenization remains pending', async () => {
+  const emitted: SubtitleData[] = [];
+  let resolveTokenization: ((value: SubtitleData | null) => void) | undefined;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async () =>
+      await new Promise<SubtitleData | null>((resolve) => {
+        resolveTokenization = resolve;
+      }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('first');
+  controller.onSubtitleChange('');
+
+  assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
+    { text: '', tokens: null },
+  ]);
+
+  assert.ok(resolveTokenization);
+  resolveTokenization({ text: 'first', tokens: [] });
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
+    { text: '', tokens: null },
+  ]);
+});
+
+test('returning to an uncached completed line emits it while another line is pending', async () => {
+  const emitted: SubtitleData[] = [];
+  let resolvePending: ((value: SubtitleData | null) => void) | undefined;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      if (text === 'A') {
+        return { text, tokens: [] };
+      }
+      return await new Promise<SubtitleData | null>((resolve) => {
+        resolvePending = resolve;
+      });
+    },
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('A');
+  await flushMicrotasks();
+  controller.invalidateTokenizationCache();
+  controller.onSubtitleChange('B');
+  controller.onSubtitleChange('A');
+
+  assert.deepEqual(emitted.at(-1), { text: 'A', tokens: null });
+  assert.ok(resolvePending);
+  resolvePending({ text: 'B', tokens: [] });
+});
+
+test('ABA subtitle changes reuse the matching first tokenization only after A is current again', async () => {
+  const emitted: SubtitleData[] = [];
+  const tokenizeCalls: string[] = [];
+  const resolvers: Array<(value: SubtitleData | null) => void> = [];
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      tokenizeCalls.push(text);
+      return await new Promise<SubtitleData | null>((resolve) => {
+        resolvers.push(resolve);
+      });
+    },
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('A');
+  controller.onSubtitleChange('B');
+  controller.onSubtitleChange('A');
+  const resolveFirst = resolvers[0];
+  assert.ok(resolveFirst);
+  resolveFirst({ text: 'A', tokens: [{ value: 1 } as never] });
+  await flushMicrotasks();
+
+  assert.deepEqual(tokenizeCalls, ['A']);
+  assert.deepEqual(emitted, [
+    { text: 'A', tokens: null },
+    { text: 'B', tokens: null },
+    { text: 'A', tokens: null },
+    { text: 'A', tokens: [{ value: 1 } as never] },
+  ]);
+});
+
+test('cached next subtitle does not downgrade to plain while processing is busy', async () => {
+  const emitted: SubtitleData[] = [];
+  let resolveTokenization: ((value: SubtitleData | null) => void) | undefined;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) =>
+      await new Promise<SubtitleData | null>((resolve) => {
+        resolveTokenization = () => resolve({ text, tokens: [] });
+      }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.preCacheTokenization('cached', { text: 'cached', tokens: [] });
+  controller.onSubtitleChange('pending');
+  controller.onSubtitleChange('cached');
+
+  assert.deepEqual(emitted, [{ text: 'pending', tokens: null }]);
+  assert.ok(resolveTokenization);
+  resolveTokenization({ text: 'pending', tokens: [] });
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: 'pending', tokens: null },
+    { text: 'cached', tokens: [] },
+  ]);
+});

src/core/services/subtitle-processing-controller.test.ts

@@ -7,18 +7,153 @@ function flushMicrotasks(): Promise<void> {
   return new Promise((resolve) => setTimeout(resolve, 0));
 }
 
-test('subtitle processing emits tokenized payload when tokenization succeeds', async () => {
+test('subtitle processing emits plain payload immediately on cache miss, then tokenized payload', async () => {
   const emitted: SubtitleData[] = [];
   const controller = createSubtitleProcessingController({
     tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
     emitSubtitle: (payload) => emitted.push(payload),
   });
 
+  controller.onSubtitleChange('字幕');
+  assert.deepEqual(emitted, [{ text: '字幕', tokens: null }]);
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: '字幕', tokens: null },
+    { text: '字幕', tokens: [] },
+  ]);
+});
+
+test('cache invalidation during pending tokenization does not re-emit the plain payload', async () => {
+  const emitted: SubtitleData[] = [];
+  const resolvers: Array<(value: SubtitleData | null) => void> = [];
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) =>
+      await new Promise<SubtitleData | null>((resolve) => {
+        resolvers.push(() => resolve({ text, tokens: [{ value: resolvers.length } as never] }));
+      }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('行');
+  assert.deepEqual(emitted, [{ text: '行', tokens: null }]);
+
+  controller.invalidateTokenizationCache();
+  resolvers[0]?.({ text: '行', tokens: [] });
+  await flushMicrotasks();
+  // Retry for the new generation is now pending; still no duplicate plain emit.
+  assert.deepEqual(emitted, [{ text: '行', tokens: null }]);
+
+  resolvers[1]?.({ text: '行', tokens: [] });
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: '行', tokens: null },
+    { text: '行', tokens: [{ value: 2 } as never] },
+  ]);
+});
+
+test('failed refresh does not downgrade an already emitted tokenized subtitle', async () => {
+  const emitted: SubtitleData[] = [];
+  let tokenizeCalls = 0;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      tokenizeCalls += 1;
+      if (tokenizeCalls > 1) {
+        throw new Error('tokenizer gone');
+      }
+      return { text, tokens: [] };
+    },
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('行');
+  await flushMicrotasks();
+  controller.invalidateTokenizationCache();
+  controller.refreshCurrentSubtitle();
+  await flushMicrotasks();
+
+  assert.equal(tokenizeCalls, 2);
+  assert.deepEqual(emitted, [
+    { text: '行', tokens: null },
+    { text: '行', tokens: [] },
+  ]);
+});
+
+test('null-tokenization refresh does not downgrade an already emitted tokenized subtitle', async () => {
+  const emitted: SubtitleData[] = [];
+  let tokenizeCalls = 0;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      tokenizeCalls += 1;
+      return tokenizeCalls > 1 ? null : { text, tokens: [] };
+    },
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('行');
+  await flushMicrotasks();
+  controller.invalidateTokenizationCache();
+  controller.refreshCurrentSubtitle();
+  await flushMicrotasks();
+
+  assert.equal(tokenizeCalls, 2);
+  assert.deepEqual(emitted, [
+    { text: '行', tokens: null },
+    { text: '行', tokens: [] },
+  ]);
+});
+
+test('subtitle processing does not emit plain payload for cached lines', async () => {
+  const emitted: SubtitleData[] = [];
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.preCacheTokenization('字幕', { text: '字幕', tokens: [] });
   controller.onSubtitleChange('字幕');
   await flushMicrotasks();
   assert.deepEqual(emitted, [{ text: '字幕', tokens: [] }]);
 });
 
+test('text that normalizes to nothing is never cached', () => {
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
+    emitSubtitle: () => {},
+  });
+
+  // Two different inputs both reduce to an empty key; sharing one entry would serve the
+  // first one's tokens for the second.
+  controller.preCacheTokenization('  ', { text: '  ', tokens: [] });
+
+  assert.equal(controller.hasCachedSubtitle('  '), false);
+  assert.equal(controller.hasCachedSubtitle('\\n'), false);
+  assert.equal(controller.consumeCachedSubtitle('\\n'), null);
+});
+
+test('subtitle processing shows plain line while tokenization is still pending', async () => {
+  const emitted: SubtitleData[] = [];
+  let resolveTokenization: ((value: SubtitleData | null) => void) | undefined;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) =>
+      await new Promise<SubtitleData | null>((resolve) => {
+        resolveTokenization = () => resolve({ text, tokens: [] });
+      }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('遅い行');
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [{ text: '遅い行', tokens: null }]);
+
+  assert.ok(resolveTokenization);
+  resolveTokenization({ text: '遅い行', tokens: [] });
+  await flushMicrotasks();
+  assert.deepEqual(emitted, [
+    { text: '遅い行', tokens: null },
+    { text: '遅い行', tokens: [] },
+  ]);
+});
+
 test('subtitle processing drops stale tokenization and delivers latest subtitle only once', async () => {
   const emitted: SubtitleData[] = [];
   let firstResolve: ((value: SubtitleData | null) => void) | undefined;
@@ -41,7 +176,11 @@ test('subtitle processing drops stale tokenization and delivers latest subtitle
   await flushMicrotasks();
   await flushMicrotasks();
 
-  assert.deepEqual(emitted, [{ text: 'second', tokens: [] }]);
+  assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
+    { text: 'second', tokens: null },
+    { text: 'second', tokens: [] },
+  ]);
 });
 
 test('subtitle processing skips duplicate subtitle emission', async () => {
@@ -60,7 +199,10 @@ test('subtitle processing skips duplicate subtitle emission', async () => {
   controller.onSubtitleChange('same');
   await flushMicrotasks();
 
-  assert.equal(emitted.length, 1);
+  assert.deepEqual(emitted, [
+    { text: 'same', tokens: null },
+    { text: 'same', tokens: [] },
+  ]);
   assert.equal(tokenizeCalls, 1);
 });
 
@@ -84,7 +226,9 @@ test('subtitle processing reuses cached tokenization for repeated subtitle text'
 
   assert.equal(tokenizeCalls, 2);
   assert.deepEqual(emitted, [
+    { text: 'first', tokens: null },
     { text: 'first', tokens: [] },
+    { text: 'second', tokens: null },
     { text: 'second', tokens: [] },
     { text: 'first', tokens: [] },
   ]);
@@ -100,7 +244,48 @@ test('subtitle processing falls back to plain subtitle when tokenization returns
   controller.onSubtitleChange('fallback');
   await flushMicrotasks();
 
+  assert.deepEqual(
+    emitted,
+    [{ text: 'fallback', tokens: null }],
+    'plain payload should not be re-emitted when tokenization yields nothing new',
+  );
+});
+
+test('null tokenization is not cached and a later cue retries tokenization', async () => {
+  const emitted: SubtitleData[] = [];
+  const callsByText = new Map<string, number>();
+  let failNext = true;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      callsByText.set(text, (callsByText.get(text) ?? 0) + 1);
+      if (text === 'fallback' && failNext) {
+        failNext = false;
+        return null;
+      }
+      return { text, tokens: [] };
+    },
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('fallback');
+  await flushMicrotasks();
+
+  assert.equal(callsByText.get('fallback'), 1);
+  assert.equal(
+    controller.hasCachedSubtitle('fallback'),
+    false,
+    'plain fallback must not be cached when tokenization yields nothing',
+  );
   assert.deepEqual(emitted, [{ text: 'fallback', tokens: null }]);
+
+  controller.onSubtitleChange('other');
+  await flushMicrotasks();
+  controller.onSubtitleChange('fallback');
+  await flushMicrotasks();
+
+  assert.equal(callsByText.get('fallback'), 2, 'later cue should retry tokenization');
+  assert.equal(controller.hasCachedSubtitle('fallback'), true);
+  assert.deepEqual(emitted.at(-1), { text: 'fallback', tokens: [] });
 });
 
 test('subtitle processing ignores duplicate current subtitle refresh without cache invalidation', async () => {
@@ -120,7 +305,10 @@ test('subtitle processing ignores duplicate current subtitle refresh without cac
   await flushMicrotasks();
 
   assert.equal(tokenizeCalls, 1);
-  assert.deepEqual(emitted, [{ text: 'same', tokens: [] }]);
+  assert.deepEqual(emitted, [
+    { text: 'same', tokens: null },
+    { text: 'same', tokens: [] },
+  ]);
 });
 
 test('subtitle processing coalesces refresh requests while current subtitle is processing', async () => {
@@ -146,7 +334,10 @@ test('subtitle processing coalesces refresh requests while current subtitle is p
   await flushMicrotasks();
 
   assert.equal(tokenizeCalls, 1);
-  assert.deepEqual(emitted, [{ text: 'same', tokens: [] }]);
+  assert.deepEqual(emitted, [
+    { text: 'same', tokens: null },
+    { text: 'same', tokens: [] },
+  ]);
 });
 
 test('subtitle processing refresh re-tokenizes after cache invalidation', async () => {
@@ -168,6 +359,7 @@ test('subtitle processing refresh re-tokenizes after cache invalidation', async
 
   assert.equal(tokenizeCalls, 2);
   assert.deepEqual(emitted, [
+    { text: 'same', tokens: null },
     { text: 'same', tokens: [{ value: 1 } as never] },
     { text: 'same', tokens: [{ value: 2 } as never] },
   ]);
@@ -183,7 +375,10 @@ test('subtitle processing refresh can use explicit text override', async () => {
   controller.refreshCurrentSubtitle('initial');
   await flushMicrotasks();
 
-  assert.deepEqual(emitted, [{ text: 'initial', tokens: [] }]);
+  assert.deepEqual(emitted, [
+    { text: 'initial', tokens: null },
+    { text: 'initial', tokens: [] },
+  ]);
 });
 
 test('subtitle processing cache invalidation only affects future subtitle events', async () => {
@@ -205,10 +400,10 @@ test('subtitle processing cache invalidation only affects future subtitle events
   await flushMicrotasks();
 
   assert.equal(callsByText.get('same'), 1);
-  assert.equal(emitted.length, 3);
+  assert.equal(emitted.length, 5);
 
   controller.invalidateTokenizationCache();
-  assert.equal(emitted.length, 3);
+  assert.equal(emitted.length, 5);
 
   controller.onSubtitleChange('different');
   await flushMicrotasks();
@@ -308,25 +503,176 @@ test('hasCachedSubtitle checks prefetched entries without consuming them', async
   assert.equal(controller.hasCachedSubtitle('猫\nです'), false);
 });
 
-test('isCacheFull returns false when cache is below limit', () => {
+test('cache keeps every entry while below the limit', () => {
   const controller = createSubtitleProcessingController({
-    tokenizeSubtitle: async (text) => ({ text, tokens: null }),
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
     emitSubtitle: () => {},
+    cacheLimit: 8,
   });
 
-  assert.equal(controller.isCacheFull(), false);
+  for (let i = 0; i < 8; i += 1) {
+    controller.preCacheTokenization(`line-${i}`, { text: `line-${i}`, tokens: [] });
+  }
+
+  assert.deepEqual(
+    Array.from({ length: 8 }, (_, i) => controller.hasCachedSubtitle(`line-${i}`)),
+    Array.from({ length: 8 }, () => true),
+  );
 });
 
-test('isCacheFull returns true when cache reaches limit', async () => {
+test('cache evicts least recently used entries once the limit is reached', () => {
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
+    emitSubtitle: () => {},
+    cacheLimit: 3,
+  });
+
+  for (const line of ['a', 'b', 'c']) {
+    controller.preCacheTokenization(line, { text: line, tokens: [] });
+  }
+  // Touching 'a' makes 'b' the eviction candidate.
+  controller.consumeCachedSubtitle('a');
+  controller.preCacheTokenization('d', { text: 'd', tokens: [] });
+
+  assert.equal(controller.hasCachedSubtitle('b'), false);
+  assert.deepEqual(
+    ['a', 'c', 'd'].map((line) => controller.hasCachedSubtitle(line)),
+    [true, true, true],
+  );
+});
+
+test('default cache limit covers a full-length title without evicting', () => {
   const controller = createSubtitleProcessingController({
     tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
     emitSubtitle: () => {},
   });
 
-  // Fill cache to the 256 limit
+  for (let i = 0; i < 2000; i += 1) {
-  for (let i = 0; i < 256; i += 1) {
     controller.preCacheTokenization(`line-${i}`, { text: `line-${i}`, tokens: [] });
   }
 
-  assert.equal(controller.isCacheFull(), true);
+  assert.equal(controller.hasCachedSubtitle('line-0'), true);
+  assert.equal(controller.hasCachedSubtitle('line-1999'), true);
+});
+
+test('onSubtitleChange reports whether processing was scheduled', async () => {
+  const emitted: SubtitleData[] = [];
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  // New text schedules work, so an emit (and anything gated on it) will follow.
+  assert.equal(controller.onSubtitleChange('字幕'), true);
+  await flushMicrotasks();
+
+  // A repeat emits nothing, so callers must not wait on an emit that is never
+  // coming (subtitle prefetching would stay paused for the rest of the cue).
+  const emittedCount = emitted.length;
+  assert.equal(controller.onSubtitleChange('字幕'), false);
+  await flushMicrotasks();
+  assert.equal(emitted.length, emittedCount);
+});
+
+test('refreshCurrentSubtitle reports the empty-text emit that an in-flight run will deliver', async () => {
+  const emitted: SubtitleData[] = [];
+  let resolveFirst: ((value: SubtitleData | null) => void) | undefined;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      if (text === '字幕') {
+        return await new Promise<SubtitleData | null>((resolve) => {
+          resolveFirst = resolve;
+        });
+      }
+      return { text, tokens: [] };
+    },
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  controller.onSubtitleChange('字幕');
+  await flushMicrotasks();
+
+  // Clearing the subtitle while tokenization is in flight: the running loop
+  // picks the empty text up and emits it, so callers gated on that emit (the
+  // prefetch pause) must be told one is coming.
+  assert.equal(controller.refreshCurrentSubtitle(''), true);
+
+  resolveFirst?.({ text: '字幕', tokens: [] });
+  await flushMicrotasks();
+  await flushMicrotasks();
+  // '字幕' is the provisional plain emit the in-flight run already made before
+  // the refresh; '' is the emit the refresh promised.
+  assert.deepEqual(
+    emitted.map((payload) => payload.text),
+    ['字幕', ''],
+  );
+});
+
+test('onProcessingSettled fires once after the queue drains, including runs that emit nothing', async () => {
+  const events: string[] = [];
+  let resolveFirst: ((value: SubtitleData | null) => void) | undefined;
+  let tokenizationFails = false;
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => {
+      if (tokenizationFails) {
+        return null;
+      }
+      if (text === '一行目') {
+        return await new Promise<SubtitleData | null>((resolve) => {
+          resolveFirst = resolve;
+        });
+      }
+      return { text, tokens: [] };
+    },
+    emitSubtitle: (payload) => events.push(`emit:${payload.text}`),
+    onProcessingSettled: () => events.push('settled'),
+  });
+
+  controller.onSubtitleChange('一行目');
+  await flushMicrotasks();
+  // A second line arrives before the first finishes: the controller still has
+  // work, so it must not report itself settled between the two.
+  controller.onSubtitleChange('二行目');
+  resolveFirst?.({ text: '一行目', tokens: [] });
+  await flushMicrotasks();
+  await flushMicrotasks();
+
+  assert.deepEqual(events, ['emit:一行目', 'emit:二行目', 'emit:二行目', 'settled']);
+
+  // Tokenization failure on a line already shown plain: nothing is emitted, and
+  // the settle signal is the only way a caller learns the work is over.
+  events.length = 0;
+  tokenizationFails = true;
+  controller.invalidateTokenizationCache();
+  assert.equal(controller.refreshCurrentSubtitle('二行目'), true);
+  await flushMicrotasks();
+  await flushMicrotasks();
+  assert.deepEqual(events, ['settled']);
+});
+
+test('notePlainSubtitleEmitted suppresses the controller repeat of a payload already shown', async () => {
+  const emitted: SubtitleData[] = [];
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  // Autoplay priming paints the plain line itself, then asks for tokenization.
+  controller.notePlainSubtitleEmitted('字幕');
+  controller.refreshCurrentSubtitle('字幕');
+  await flushMicrotasks();
+
+  assert.deepEqual(emitted, [{ text: '字幕', tokens: [] }]);
+});
+
+test('refreshCurrentSubtitle reports no emit for empty text when nothing is running', async () => {
+  const emitted: SubtitleData[] = [];
+  const controller = createSubtitleProcessingController({
+    tokenizeSubtitle: async (text) => ({ text, tokens: [] }),
+    emitSubtitle: (payload) => emitted.push(payload),
+  });
+
+  assert.equal(controller.refreshCurrentSubtitle(''), false);
+  await flushMicrotasks();
+  assert.deepEqual(emitted, []);
 });

src/core/services/subtitle-processing-controller.ts

@@ -1,32 +1,76 @@
 import type { SubtitleData } from '../../types';
+import { normalizePlainSubtitleText } from './ass-text';
 
 export interface SubtitleProcessingControllerDeps {
   tokenizeSubtitle: (text: string) => Promise<SubtitleData | null>;
   emitSubtitle: (payload: SubtitleData) => void;
+  /**
+   * Fires when the controller runs out of work: every scheduled line has been
+   * processed, whether it ended in an emit, a suppressed duplicate, or a
+   * tokenizer failure. Callers that hold a resource for the duration of
+   * processing (prefetch pausing) release it here rather than on an emit,
+   * which is not guaranteed to happen.
+   */
+  onProcessingSettled?: () => void;
   logDebug?: (message: string) => void;
   now?: () => number;
+  cacheLimit?: number;
 }
 
+/**
+ * Pure memory bound on the LRU, not a coverage limit: prefetching runs to the end of a
+ * file regardless of cache pressure. Sized to hold a feature-length title (a 24-minute
+ * episode runs 300-400 lines, a 2-hour film ~2000) plus room for lines that repeat across
+ * episodes of a series, so openings and endings stay warm between titles.
+ */
+export const DEFAULT_SUBTITLE_TOKENIZATION_CACHE_LIMIT = 2500;
+
 export interface SubtitleProcessingController {
-  onSubtitleChange: (text: string) => void;
+  /**
-  refreshCurrentSubtitle: (textOverride?: string) => void;
+   * Returns whether processing is now scheduled or already in flight for this
+   * event. A false return means the controller is idle and will do nothing, so
+   * onProcessingSettled will not fire; callers that pause work for the duration
+   * of processing (such as subtitle prefetching) must release it themselves.
+   */
+  onSubtitleChange: (text: string) => boolean;
+  /** Same contract as onSubtitleChange: whether processing is pending. */
+  refreshCurrentSubtitle: (textOverride?: string) => boolean;
+  /**
+   * Records that this exact text has already been shown plain by someone else
+   * (autoplay priming paints its first frame before scheduling tokenization),
+   * so the controller does not repeat that payload on its way to the tokenized
+   * one.
+   */
+  notePlainSubtitleEmitted: (text: string) => void;
   invalidateTokenizationCache: () => void;
   preCacheTokenization: (text: string, data: SubtitleData) => void;
   consumeCachedSubtitle: (text: string) => SubtitleData | null;
   hasCachedSubtitle: (text: string) => boolean;
-  isCacheFull: () => boolean;
 }
 
+/**
+ * Prefetched cues and live mpv text are both already decoded from ASS, so the key only
+ * has to settle whitespace for one authored line to resolve to one entry.
+ *
+ * An empty key is not a line: it is whatever normalization reduced to nothing. Callers
+ * must skip the cache for it rather than let every such input share one entry.
+ */
 export function normalizeSubtitleCacheKey(text: string): string {
-  return text.replace(/\r\n/g, '\n').replace(/\\N/g, '\n').replace(/\\n/g, '\n').trim();
+  return normalizePlainSubtitleText(text);
 }
 
 export function createSubtitleProcessingController(
   deps: SubtitleProcessingControllerDeps,
 ): SubtitleProcessingController {
-  const SUBTITLE_TOKENIZATION_CACHE_LIMIT = 256;
+  const SUBTITLE_TOKENIZATION_CACHE_LIMIT =
+    deps.cacheLimit && deps.cacheLimit > 0
+      ? deps.cacheLimit
+      : DEFAULT_SUBTITLE_TOKENIZATION_CACHE_LIMIT;
   let latestText = '';
   let lastEmittedText = '';
+  // Tracks the latest provisional plain emit across rapid changes and loop retries
+  // so the same line is never shown plain twice.
+  let lastPlainEmittedText: string | null = null;
   let cacheGeneration = 0;
   let lastEmittedGeneration = 0;
   let processing = false;
@@ -36,6 +80,9 @@ export function createSubtitleProcessingController(
 
   const getCachedTokenization = (text: string): SubtitleData | null => {
     const cacheKey = normalizeSubtitleCacheKey(text);
+    if (!cacheKey) {
+      return null;
+    }
     const cached = tokenizationCache.get(cacheKey);
     if (!cached) {
       return null;
@@ -47,7 +94,11 @@ export function createSubtitleProcessingController(
   };
 
   const setCachedTokenization = (text: string, payload: SubtitleData): void => {
-    tokenizationCache.set(normalizeSubtitleCacheKey(text), payload);
+    const cacheKey = normalizeSubtitleCacheKey(text);
+    if (!cacheKey) {
+      return;
+    }
+    tokenizationCache.set(cacheKey, payload);
     while (tokenizationCache.size > SUBTITLE_TOKENIZATION_CACHE_LIMIT) {
       const firstKey = tokenizationCache.keys().next().value;
       if (firstKey !== undefined) {
@@ -70,9 +121,12 @@ export function createSubtitleProcessingController(
         const startedAtMs = now();
 
         if (!text.trim()) {
-          deps.emitSubtitle({ text, tokens: null });
+          if (lastPlainEmittedText !== text) {
+            deps.emitSubtitle({ text, tokens: null });
+          }
           lastEmittedText = text;
           lastEmittedGeneration = generation;
+          lastPlainEmittedText = null;
           break;
         }
 
@@ -82,11 +136,25 @@ export function createSubtitleProcessingController(
           if (cachedTokenized) {
             output = cachedTokenized;
           } else {
+            // Cache miss: show the plain line on time; the tokenized payload
+            // upgrades it once ready. Skipped on refreshes of an already
+            // emitted line so downstream consumers never see a downgrade.
+            if (text !== lastEmittedText && text !== lastPlainEmittedText) {
+              deps.emitSubtitle({ text, tokens: null });
+              lastPlainEmittedText = text;
+            }
             const tokenized = await deps.tokenizeSubtitle(text);
+            // A null result is a transient tokenizer failure, not a verdict on
+            // the line: caching the plain fallback would pin it untokenized for
+            // every later occurrence.
             if (tokenized) {
               output = tokenized;
+              // A result computed before an invalidation must not repopulate the
+              // fresh cache, or the retry below would serve the stale entry.
+              if (generation === cacheGeneration) {
+                setCachedTokenization(text, tokenized);
+              }
             }
-            setCachedTokenization(text, output);
           }
         } catch (error) {
           deps.logDebug?.(`Subtitle tokenization failed: ${(error as Error).message}`);
@@ -107,9 +175,16 @@ export function createSubtitleProcessingController(
           continue;
         }
 
-        deps.emitSubtitle(output);
+        // An untokenized result adds nothing when this line was already shown,
+        // either provisionally or as an earlier full emit (failed refresh) —
+        // emitting it would duplicate or downgrade what is on screen.
+        const plainAlreadyShown = lastPlainEmittedText === text || lastEmittedText === text;
+        if (!(output.tokens === null && output.text === text && plainAlreadyShown)) {
+          deps.emitSubtitle(output);
+        }
         lastEmittedText = text;
         lastEmittedGeneration = generation;
+        lastPlainEmittedText = null;
         deps.logDebug?.(
           `Subtitle tokenization delivered; elapsed=${now() - startedAtMs}ms, staleDrops=${staleDropCount}`,
         );
@@ -126,32 +201,53 @@ export function createSubtitleProcessingController(
           (latestText.trim() && cacheGeneration !== lastEmittedGeneration)
         ) {
           processLatest();
+          return;
         }
+        // Nothing left to do: signal completion even when this run emitted
+        // nothing (suppressed duplicate, tokenizer failure), or callers waiting
+        // on the controller would wait forever.
+        deps.onProcessingSettled?.();
       });
   };
 
   return {
     onSubtitleChange: (text: string) => {
       if (text === latestText) {
-        return;
+        // A run already in flight for this text will still emit for it.
+        return processing;
       }
       latestText = text;
+      if (
+        processing &&
+        text !== lastPlainEmittedText &&
+        !tokenizationCache.has(normalizeSubtitleCacheKey(text))
+      ) {
+        deps.emitSubtitle({ text, tokens: null });
+        lastPlainEmittedText = text;
+      }
       processLatest();
+      return true;
     },
     refreshCurrentSubtitle: (textOverride?: string) => {
       if (typeof textOverride === 'string') {
         latestText = textOverride;
       }
       if (!latestText.trim()) {
-        return;
+        // A run in flight will pick this up and emit the empty subtitle, so
+        // the caller is still waiting on an emit.
+        return processing;
       }
-      if (
+      if (processing) {
-        processing ||
+        return true;
-        (latestText === lastEmittedText && cacheGeneration === lastEmittedGeneration)
+      }
-      ) {
+      if (latestText === lastEmittedText && cacheGeneration === lastEmittedGeneration) {
-        return;
+        return false;
       }
       processLatest();
+      return true;
+    },
+    notePlainSubtitleEmitted: (text: string) => {
+      lastPlainEmittedText = text;
     },
     invalidateTokenizationCache: () => {
       tokenizationCache.clear();
@@ -169,13 +265,12 @@ export function createSubtitleProcessingController(
       latestText = text;
       lastEmittedText = text;
       lastEmittedGeneration = cacheGeneration;
+      lastPlainEmittedText = null;
       return cached;
     },
     hasCachedSubtitle: (text: string) => {
-      return tokenizationCache.has(normalizeSubtitleCacheKey(text));
+      const cacheKey = normalizeSubtitleCacheKey(text);
-    },
+      return cacheKey.length > 0 && tokenizationCache.has(cacheKey);
-    isCacheFull: () => {
-      return tokenizationCache.size >= SUBTITLE_TOKENIZATION_CACHE_LIMIT;
     },
   };
 }

src/core/services/subtitle-ws.test.ts

@@ -1,6 +1,7 @@
 import test from 'node:test';
 import assert from 'node:assert/strict';
 import {
+  isSubtitleAnnotationUpgrade,
   serializeInitialSubtitleWebsocketMessage,
   serializeSubtitleMarkup,
   serializeSubtitleWebsocketMessage,
@@ -13,6 +14,40 @@ const frequencyOptions = {
   mode: 'banded' as const,
 };
 
+test('annotation upgrade requires matching text and cue timing', () => {
+  const current: SubtitleData = {
+    text: '字幕',
+    tokens: null,
+    startTime: 10,
+    endTime: 12,
+  };
+
+  assert.equal(
+    isSubtitleAnnotationUpgrade(current, {
+      ...current,
+      tokens: [],
+    }),
+    true,
+  );
+  assert.equal(
+    isSubtitleAnnotationUpgrade(current, {
+      ...current,
+      tokens: [],
+      startTime: 11,
+    }),
+    false,
+  );
+  assert.equal(
+    isSubtitleAnnotationUpgrade(current, {
+      ...current,
+      text: '次の字幕',
+      tokens: [],
+    }),
+    false,
+  );
+  assert.equal(isSubtitleAnnotationUpgrade(current, current), false);
+});
+
 test('serializeSubtitleMarkup escapes plain text and preserves line breaks', () => {
   const payload: SubtitleData = {
     text: 'a < b\nx & y',

src/core/services/subtitle-ws.ts

@@ -20,6 +20,20 @@ export type SubtitleWebsocketFrequencyOptions = {
 
 export type SubtitleWebsocketPayloadMode = 'plain' | 'annotated';
 
+export function isSubtitleAnnotationUpgrade(
+  current: SubtitleData | null,
+  next: SubtitleData,
+): boolean {
+  return (
+    current !== null &&
+    current.tokens === null &&
+    next.tokens !== null &&
+    current.text === next.text &&
+    current.startTime === next.startTime &&
+    current.endTime === next.endTime
+  );
+}
+
 type SubtitleWebsocketMessageOptions = {
   payloadMode?: SubtitleWebsocketPayloadMode;
 };

src/core/services/tokenizer.test.ts

@@ -1651,9 +1651,11 @@ test('tokenizeSubtitle clears JLPT level from standalone Yomitan particle token'
   assert.equal(result.tokens?.[0]?.jlptLevel, undefined);
 });
 
-test('tokenizeSubtitle returns null tokens for empty normalized text', async () => {
+test('tokenizeSubtitle returns the normalized text when it comes out empty', async () => {
+  // Handing back the original would push whatever normalization dropped into app state
+  // as if it were subtitle text.
   const result = await tokenizeSubtitle(' \\n  ', makeDeps());
-  assert.deepEqual(result, { text: ' \\n  ', tokens: null });
+  assert.deepEqual(result, { text: '', tokens: null });
 });
 
 test('tokenizeSubtitle normalizes newlines before Yomitan parse request', async () => {
@@ -2934,44 +2936,12 @@ test('tokenizeSubtitle preserves Yomitan compound token when MeCab components ar
                 return [];
               }
 
-              if (script.includes('parseText')) {
-                return [
-                  {
-                    source: 'scanning-parser',
-                    index: 0,
-                    content: [
-                      [
-                        {
-                          text: '取り組んで',
-                          reading: 'とりくんで',
-                          headwords: [[{ term: '取り組む' }]],
-                        },
-                      ],
-                      [
-                        {
-                          text: 'もらいます',
-                          reading: 'もらいます',
-                          headwords: [[{ term: 'もらう' }]],
-                        },
-                      ],
-                    ],
-                  },
-                ];
-              }
-
               return [
                 {
-                  surface: '取り',
+                  surface: '取り組んで',
-                  reading: 'とり',
+                  reading: 'とりくんで',
-                  headword: '取る',
+                  headword: '取り組む',
                   startPos: 0,
-                  endPos: 2,
-                },
-                {
-                  surface: '組んで',
-                  reading: 'くんで',
-                  headword: '組む',
-                  startPos: 2,
                   endPos: 5,
                 },
                 {

src/core/services/tokenizer.ts

@@ -27,6 +27,7 @@ import {
 } from './tokenizer/yomitan-parser-runtime';
 import type { YomitanTermFrequency } from './tokenizer/yomitan-parser-runtime';
 import { isKanaChar } from './tokenizer/token-classification';
+import { normalizePlainSubtitleText } from './ass-text';
 
 const logger = createLogger('main:tokenizer');
 
@@ -70,6 +71,7 @@ export interface TokenizerServiceDeps {
   getNameMatchImagesEnabled?: () => boolean;
   getCharacterNameImage?: (term: string) => CharacterNameImage | null;
   getCurrentCharacterDictionaryMediaId?: () => number | null;
+  getCharacterNameCandidates?: () => { key: string; forms: string[] } | null;
   getFrequencyDictionaryEnabled?: () => boolean;
   getFrequencyDictionaryMatchMode?: () => FrequencyDictionaryMatchMode;
   getFrequencyRank?: FrequencyDictionaryLookup;
@@ -106,6 +108,7 @@ export interface TokenizerDepsRuntimeOptions {
   getNameMatchImagesEnabled?: () => boolean;
   getCharacterNameImage?: (term: string) => CharacterNameImage | null;
   getCurrentCharacterDictionaryMediaId?: () => number | null;
+  getCharacterNameCandidates?: () => { key: string; forms: string[] } | null;
   getFrequencyDictionaryEnabled?: () => boolean;
   getFrequencyDictionaryMatchMode?: () => FrequencyDictionaryMatchMode;
   getFrequencyRank?: FrequencyDictionaryLookup;
@@ -266,6 +269,7 @@ export function createTokenizerDepsRuntime(
     getNameMatchImagesEnabled: options.getNameMatchImagesEnabled,
     getCharacterNameImage: options.getCharacterNameImage,
     getCurrentCharacterDictionaryMediaId: options.getCurrentCharacterDictionaryMediaId,
+    getCharacterNameCandidates: options.getCharacterNameCandidates,
     getFrequencyDictionaryEnabled: options.getFrequencyDictionaryEnabled,
     getFrequencyDictionaryMatchMode: options.getFrequencyDictionaryMatchMode ?? (() => 'headword'),
     getFrequencyRank: options.getFrequencyRank,
@@ -716,15 +720,30 @@ function getAnnotationOptions(deps: TokenizerServiceDeps): TokenizerAnnotationOp
   };
 }
 
+// Per-line stage durations for the pipeline debug log; every field is filled in
+// by the stage that awaits the corresponding work.
+interface TokenizationStageTimings {
+  scanMs?: number;
+  mecabMs?: number;
+  frequencyMs?: number;
+  annotateMs?: number;
+}
+
 async function parseWithYomitanInternalParser(
   text: string,
   deps: TokenizerServiceDeps,
   options: TokenizerAnnotationOptions,
+  stageTimings?: TokenizationStageTimings,
 ): Promise<MergedToken[] | null> {
+  const scanStartedAtMs = Date.now();
   const selectedTokens = await requestYomitanScanTokens(text, deps, logger, {
     includeNameMatchMetadata: options.nameMatchEnabled,
     currentCharacterDictionaryMediaId: deps.getCurrentCharacterDictionaryMediaId?.() ?? null,
+    nameCandidates: deps.getCharacterNameCandidates?.() ?? null,
   });
+  if (stageTimings) {
+    stageTimings.scanMs = Date.now() - scanStartedAtMs;
+  }
   if (!selectedTokens || selectedTokens.length === 0) {
     return null;
   }
@@ -757,6 +776,7 @@ async function parseWithYomitanInternalParser(
 
   const frequencyRankPromise: Promise<YomitanFrequencyIndex> = options.frequencyEnabled
     ? (async () => {
+        const frequencyStartedAtMs = Date.now();
         const frequencyMatchMode = options.frequencyMatchMode;
         const termReadingList = buildYomitanFrequencyTermReadingList(
           normalizedSelectedTokens,
@@ -767,12 +787,17 @@ async function parseWithYomitanInternalParser(
           deps,
           logger,
         );
-        return buildYomitanFrequencyIndex(yomitanFrequencies);
+        const frequencyIndex = buildYomitanFrequencyIndex(yomitanFrequencies);
+        if (stageTimings) {
+          stageTimings.frequencyMs = Date.now() - frequencyStartedAtMs;
+        }
+        return frequencyIndex;
       })()
     : Promise.resolve({ byPair: new Map(), byTerm: new Map() });
 
   const mecabEnrichmentPromise: Promise<MergedToken[]> = needsMecabPosEnrichment(options)
     ? (async () => {
+        const mecabStartedAtMs = Date.now();
         try {
           const mecabTokens = await deps.tokenizeWithMecab(text);
           const enrichTokensWithMecab = deps.enrichTokensWithMecab ?? enrichTokensWithMecabAsync;
@@ -786,6 +811,10 @@ async function parseWithYomitanInternalParser(
             `textLength=${text.length}`,
           );
           return normalizedSelectedTokens;
+        } finally {
+          if (stageTimings) {
+            stageTimings.mecabMs = Date.now() - mecabStartedAtMs;
+          }
         }
       })()
     : Promise.resolve(normalizedSelectedTokens);
@@ -858,14 +887,14 @@ export async function tokenizeSubtitle(
   text: string,
   deps: TokenizerServiceDeps,
 ): Promise<SubtitleData> {
-  const displayText = text
+  const displayText = normalizePlainSubtitleText(text);
-    .replace(/\r\n/g, '\n')
-    .replace(/\\N/g, '\n')
-    .replace(/\\n/g, '\n')
-    .trim();
 
+  // ASS decoding already happened upstream (cue parser for files, mpv for live text), so
+  // all this drops is whitespace -- but a whitespace-only line still normalizes to empty.
+  // Return the normalized form anyway: handing back the original would put a blank line
+  // into application state as if it were subtitle text.
   if (!displayText) {
-    return { text, tokens: null };
+    return { text: displayText, tokens: null };
   }
 
   const tokenizeText = displayText
@@ -876,15 +905,35 @@ export async function tokenizeSubtitle(
   const annotationOptions = getAnnotationOptions(deps);
   annotationOptions.sourceText = tokenizeText;
 
-  const yomitanTokens = await parseWithYomitanInternalParser(tokenizeText, deps, annotationOptions);
+  const stageTimings: TokenizationStageTimings = {};
+  const startedAtMs = Date.now();
+  const logStageTimings = (tokenCount: number): void => {
+    logger.debug(
+      `Subtitle tokenization stages; textLength=${tokenizeText.length}, tokenCount=${tokenCount}, ` +
+        `scanMs=${stageTimings.scanMs ?? '-'}, mecabMs=${stageTimings.mecabMs ?? '-'}, ` +
+        `frequencyMs=${stageTimings.frequencyMs ?? '-'}, annotateMs=${stageTimings.annotateMs ?? '-'}, ` +
+        `totalMs=${Date.now() - startedAtMs}`,
+    );
+  };
+
+  const yomitanTokens = await parseWithYomitanInternalParser(
+    tokenizeText,
+    deps,
+    annotationOptions,
+    stageTimings,
+  );
   if (yomitanTokens && yomitanTokens.length > 0) {
+    const annotateStartedAtMs = Date.now();
     const annotatedTokens = await applyAnnotationStage(yomitanTokens, deps, annotationOptions);
+    stageTimings.annotateMs = Date.now() - annotateStartedAtMs;
     const renderedTokens = applyCharacterNameImages(annotatedTokens, deps, annotationOptions);
+    logStageTimings(renderedTokens.length);
     return {
       text: displayText,
       tokens: renderedTokens.length > 0 ? renderedTokens : null,
     };
   }
 
+  logStageTimings(0);
   return { text: displayText, tokens: null };
 }

src/core/services/tokenizer/character-dictionary-title.ts

@@ -0,0 +1,4 @@
+// Title prefix of the dictionaries SubMiner generates per media. Lives on its
+// own because both the main process and the injected scan runtime match on it,
+// and the injected fragments interpolate it into their own source.
+export const CHARACTER_DICTIONARY_TITLE_PREFIX = 'SubMiner Character Dictionary';

src/core/services/tokenizer/golden-corpus-harness.ts

@@ -366,8 +366,11 @@ export function createReplayMessageStore(messages: GoldenRecordedMessage[]): Rep
   };
 }
 
-async function runInjectedScriptInVm(script: string, store: ReplayMessageStore): Promise<unknown> {
+// One persistent context per fixture, matching the real parser window: the
-  return await vm.runInNewContext(script, {
+// scan runtime installs itself once into globalThis and later per-line call
+// scripts reuse it.
+function createInjectedScriptVm(store: ReplayMessageStore): (script: string) => Promise<unknown> {
+  const context = vm.createContext({
     chrome: {
       runtime: {
         lastError: null,
@@ -393,6 +396,7 @@ async function runInjectedScriptInVm(script: string, store: ReplayMessageStore):
     Set,
     String,
   });
+  return async (script: string) => await vm.runInContext(script, context);
 }
 
 export function createReplayTokenizerDeps(fixture: GoldenFixture): TokenizerServiceDeps {
@@ -400,13 +404,14 @@ export function createReplayTokenizerDeps(fixture: GoldenFixture): TokenizerServ
   const scriptResults = new Map(
     fixture.recording.scripts.map((entry) => [entry.sha256, entry] as const),
   );
+  const runInjectedScriptInVm = createInjectedScriptVm(store);
 
   const parserWindow = {
     isDestroyed: () => false,
     webContents: {
       executeJavaScript: async (script: string) => {
         try {
-          return await runInjectedScriptInVm(script, store);
+          return await runInjectedScriptInVm(script);
         } catch (vmError) {
           const recorded = scriptResults.get(hashInjectedScript(script));
           if (recorded) {

src/core/services/tokenizer/token-classification.test.ts

@@ -8,6 +8,7 @@ import {
   isKanaChar,
   isKanaOnlyText,
   isTokenPos2Excluded,
+  normalizeKana,
 } from './token-classification';
 
 const POS1_EXCLUSIONS = new Set(['助詞']);
@@ -29,6 +30,26 @@ function makeNoun(surface: string): MergedToken {
   };
 }
 
+test('kana normalization folds halfwidth kana, composing the voiced pairs', () => {
+  // ｶ + ﾞ is two code points for one character: without composing them, a
+  // halfwidth word counts as longer than the reading that spells it, which
+  // disqualifies the reading from known-word matching.
+  assert.equal(normalizeKana('ｶﾞｸ'), normalizeKana('ガク'));
+  assert.equal(normalizeKana('ﾊﾟﾝ'), normalizeKana('パン'));
+  assert.equal(normalizeKana('ﾐﾅﾄ'), 'みなと');
+  assert.ok(isKanaOnlyText('ｶﾞｸ'));
+});
+
+test('kana normalization leaves characters other than halfwidth kana alone', () => {
+  // The composition is scoped to the halfwidth runs: applied to the whole
+  // string, NFKC would also rewrite these into something the dictionary, the
+  // known-word list, and the frequency data were never keyed on.
+  assert.equal(normalizeKana('①ｶﾞ'), '①が');
+  assert.equal(normalizeKana('Ａｶﾞ'), 'Ａが');
+  assert.equal(normalizeKana('㍑ｶﾞ'), '㍑が');
+  assert.equal(normalizeKana('fiｶﾞ'), 'fiが');
+});
+
 test('kana classification excludes the katakana-hiragana double hyphen', () => {
   assert.equal(isKanaChar('゠'), false);
   assert.equal(isKanaOnlyText('゠'), false);

src/core/services/tokenizer/token-classification.ts

@@ -4,8 +4,20 @@ const KATAKANA_TO_HIRAGANA_OFFSET = 0x60;
 const KATAKANA_CODEPOINT_START = 0x30a1;
 const KATAKANA_CODEPOINT_END = 0x30f6;
 
+// No `u` flag: the range is entirely BMP so it changes nothing here, and
+// Bun's unicode-mode matcher mis-handles this class next to certain ligatures.
+const HALFWIDTH_KANA_RUN = /[\uff66-\uff9f]+/g;
+
+// NFKC over the halfwidth kana only, never the whole string: it composes the
+// voiced pairs (ｶ + ﾞ) into single characters so ｶﾞｸ compares equal to ガク
+// instead of counting one character longer than the word it spells, but run
+// over everything it would also rewrite unrelated text (① → 1, ㍑ → リットル).
+function composeHalfwidthKana(text: string): string {
+  return text.replace(HALFWIDTH_KANA_RUN, (run) => run.normalize('NFKC'));
+}
+
 export function normalizeKana(text: string): string {
-  const raw = text.trim();
+  const raw = composeHalfwidthKana(text).trim();
   if (!raw) {
     return '';
   }

src/core/services/tokenizer/yomitan-dictionary-classification-script.ts

@@ -0,0 +1,150 @@
+// Dictionary classification for the injected scan runtime: which dictionaries
+// an entry came from, and whether it is a SubMiner character entry for the
+// media being watched. Both walk nested entry data, so both are memoized on the
+// entry object by the runtime that hosts them.
+
+import { CHARACTER_DICTIONARY_TITLE_PREFIX } from './character-dictionary-title';
+
+// The prefix is interpolated into generated regex source, so metacharacters in
+// it would change what the pattern matches (or fail to compile).
+const ESCAPED_TITLE_PREFIX = CHARACTER_DICTIONARY_TITLE_PREFIX.replace(
+  /[.*+?^${}()|[\]\\]/g,
+  '\\$&',
+);
+const TITLE_MEDIA_ID_PATTERN = ESCAPED_TITLE_PREFIX + String.raw`[^\d]*(?:AniList\s*)?(\d+)`;
+
+export const YOMITAN_DICTIONARY_CLASSIFICATION_HELPERS = String.raw`
+      function normalizeWordClasses(headword) {
+          if (!Array.isArray(headword?.wordClasses)) { return undefined; }
+          const classes = headword.wordClasses.filter((wordClass) => typeof wordClass === "string" && wordClass.trim().length > 0);
+          return classes.length > 0 ? classes : undefined;
+        }
+        function appendDictionaryNames(target, value) {
+          if (!value || typeof value !== 'object') {
+            return;
+          }
+          const candidates = [
+            value.dictionary,
+            value.dictionaryName,
+            value.name,
+            value.title,
+            value.dictionaryTitle,
+            value.dictionaryAlias
+          ];
+          for (const candidate of candidates) {
+            if (typeof candidate === 'string' && candidate.trim().length > 0) {
+              target.push(candidate.trim());
+            }
+          }
+        }
+        // Memoized on the entry object: termsFind results are cached across
+        // lines, so the same entries come back for every repeated lookup, and
+        // each one is classified several times per scan (name pre-pass,
+        // headword preference, every retry window).
+        function getDictionaryEntryNames(entry) {
+          if (!entry || typeof entry !== 'object') { return []; }
+          const cached = dictionaryEntryNamesCache.get(entry);
+          if (cached !== undefined) { return cached; }
+          const names = [];
+          appendDictionaryNames(names, entry);
+          for (const definition of entry?.definitions || []) {
+            appendDictionaryNames(names, definition);
+          }
+          for (const frequency of entry?.frequencies || []) {
+            appendDictionaryNames(names, frequency);
+          }
+          for (const pronunciation of entry?.pronunciations || []) {
+            appendDictionaryNames(names, pronunciation);
+          }
+          dictionaryEntryNamesCache.set(entry, names);
+          return names;
+        }
+        // Cached per scan rather than per runtime: the answer depends on
+        // includeNameMatchMetadata, which is a per-call parameter.
+        const nameDictionaryEntryCache = new WeakMap();
+        function isNameDictionaryEntry(entry) {
+          if (!includeNameMatchMetadata || !entry || typeof entry !== 'object') {
+            return false;
+          }
+          const cached = nameDictionaryEntryCache.get(entry);
+          if (cached !== undefined) { return cached; }
+          const isName = getDictionaryEntryNames(entry).some((name) => name.startsWith(${JSON.stringify(CHARACTER_DICTIONARY_TITLE_PREFIX)}));
+          nameDictionaryEntryCache.set(entry, isName);
+          return isName;
+        }
+        const TITLE_MEDIA_ID_REGEX = new RegExp(${JSON.stringify(TITLE_MEDIA_ID_PATTERN)}, 'i');
+        function parseSubMinerMediaIdFromString(value) {
+          const imageMatch = value.match(/\bimg\/m(\d+)-/i);
+          if (imageMatch) {
+            const parsed = Number.parseInt(imageMatch[1], 10);
+            if (Number.isSafeInteger(parsed) && parsed > 0) { return parsed; }
+          }
+          const titleMatch = value.match(TITLE_MEDIA_ID_REGEX);
+          if (titleMatch) {
+            const parsed = Number.parseInt(titleMatch[1], 10);
+            if (Number.isSafeInteger(parsed) && parsed > 0) { return parsed; }
+          }
+          return null;
+        }
+        function parseSubMinerMediaIdCandidate(value) {
+          if (typeof value === 'number' && Number.isSafeInteger(value) && value > 0) {
+            return value;
+          }
+          if (typeof value === 'string' && /^\d+$/.test(value.trim())) {
+            const parsed = Number.parseInt(value.trim(), 10);
+            if (Number.isSafeInteger(parsed) && parsed > 0) { return parsed; }
+          }
+          return null;
+        }
+        function collectSubMinerMediaIds(value, target) {
+          if (typeof value === 'string') {
+            const parsed = parseSubMinerMediaIdFromString(value);
+            if (parsed !== null) { target.add(parsed); }
+            return;
+          }
+          if (!value || typeof value !== 'object') {
+            return;
+          }
+          if (Array.isArray(value)) {
+            for (const item of value) { collectSubMinerMediaIds(item, target); }
+            return;
+          }
+          const mediaIdCandidates = [
+            value.subminerMediaId,
+            value.subMinerMediaId,
+            value.characterDictionaryMediaId,
+            value.data?.subminerMediaId,
+            value.data?.subMinerMediaId,
+            value.data?.characterDictionaryMediaId
+          ];
+          for (const candidate of mediaIdCandidates) {
+            const parsed = parseSubMinerMediaIdCandidate(candidate);
+            if (parsed !== null) { target.add(parsed); }
+          }
+          for (const child of Object.values(value)) {
+            collectSubMinerMediaIds(child, target);
+          }
+        }
+        // Walking an entry collects media ids from every nested value, so this
+        // is the most expensive classification step; memoized on the entry for
+        // the same reason as the dictionary names above.
+        function getSubMinerMediaIds(entry) {
+          if (!entry || typeof entry !== 'object') { return EMPTY_MEDIA_ID_SET; }
+          const cached = subMinerMediaIdsCache.get(entry);
+          if (cached !== undefined) { return cached; }
+          const mediaIds = new Set();
+          collectSubMinerMediaIds(entry, mediaIds);
+          subMinerMediaIdsCache.set(entry, mediaIds);
+          return mediaIds;
+        }
+        function isCurrentMediaNameDictionaryEntry(entry) {
+          if (!isNameDictionaryEntry(entry)) {
+            return false;
+          }
+          if (currentCharacterDictionaryMediaId === null) {
+            return true;
+          }
+          const mediaIds = getSubMinerMediaIds(entry);
+          return mediaIds.size === 0 || mediaIds.has(currentCharacterDictionaryMediaId);
+      }
+`;

src/core/services/tokenizer/yomitan-frequency-script.ts

@@ -0,0 +1,135 @@
+// Frequency-rank resolution for the injected scan runtime: reads the many
+// shapes a Yomitan frequency entry can take and picks the best rank for a
+// headword, honouring per-dictionary priority and occurrence-vs-rank mode.
+
+export const YOMITAN_FREQUENCY_HELPERS = String.raw`
+      function parsePositiveFrequencyNumber(value) {
+        if (typeof value === 'number' && Number.isFinite(value) && value > 0) {
+          return Math.max(1, Math.floor(value));
+        }
+        if (typeof value === 'string') {
+          const numericMatch = value.trim().match(/[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?/)?.[0];
+          if (!numericMatch) { return null; }
+          const parsed = Number.parseFloat(numericMatch);
+          if (!Number.isFinite(parsed) || parsed <= 0) { return null; }
+          return Math.max(1, Math.floor(parsed));
+        }
+        if (Array.isArray(value)) {
+          for (const item of value) {
+            const parsed = parsePositiveFrequencyNumber(item);
+            if (parsed !== null) { return parsed; }
+          }
+        }
+        return null;
+      }
+      function parseDisplayFrequencyNumber(value) {
+        if (typeof value === 'string') {
+          const leadingDigits = value.trim().match(/^\d+/)?.[0];
+          if (!leadingDigits) { return null; }
+          const parsed = Number.parseInt(leadingDigits, 10);
+          return Number.isFinite(parsed) && parsed > 0 ? parsed : null;
+        }
+        return parsePositiveFrequencyNumber(value);
+      }
+      function getFrequencyDictionaryName(frequency) {
+        const candidates = [
+          frequency?.dictionary,
+          frequency?.dictionaryName,
+          frequency?.name,
+          frequency?.title,
+          frequency?.dictionaryTitle,
+          frequency?.dictionaryAlias
+        ];
+        for (const candidate of candidates) {
+          if (typeof candidate === 'string' && candidate.trim().length > 0) {
+            return candidate.trim();
+          }
+        }
+        return null;
+      }
+      function getBestFrequencyRank(dictionaryEntry, headwordIndex, dictionaryPriorityByName, dictionaryFrequencyModeByName) {
+        let best = null;
+        const headwordCount = Array.isArray(dictionaryEntry?.headwords) ? dictionaryEntry.headwords.length : 0;
+        for (const frequency of dictionaryEntry?.frequencies || []) {
+          if (!frequency || typeof frequency !== 'object') { continue; }
+          const frequencyHeadwordIndex = frequency.headwordIndex;
+          if (typeof frequencyHeadwordIndex === 'number') {
+            if (frequencyHeadwordIndex !== headwordIndex) { continue; }
+          } else if (headwordCount > 1) {
+            continue;
+          }
+          const dictionary = getFrequencyDictionaryName(frequency);
+          if (!dictionary) { continue; }
+          if (dictionaryFrequencyModeByName[dictionary] === 'occurrence-based') { continue; }
+          const rank =
+            parseDisplayFrequencyNumber(frequency.displayValue) ??
+            parsePositiveFrequencyNumber(frequency.frequency);
+          if (rank === null) { continue; }
+          const priorityRaw = dictionaryPriorityByName[dictionary];
+          const fallbackPriority =
+            typeof frequency.dictionaryIndex === 'number' && Number.isFinite(frequency.dictionaryIndex)
+              ? Math.max(0, Math.floor(frequency.dictionaryIndex))
+              : Number.MAX_SAFE_INTEGER;
+          const priority =
+            typeof priorityRaw === 'number' && Number.isFinite(priorityRaw)
+              ? Math.max(0, Math.floor(priorityRaw))
+              : fallbackPriority;
+          if (best === null || priority < best.priority || (priority === best.priority && rank < best.rank)) {
+            best = { priority, rank };
+          }
+        }
+        return best?.rank ?? null;
+      }
+      function hasExactSource(headword, token, requirePrimary) {
+        for (const src of headword?.sources || []) {
+          if (src.originalText !== token) { continue; }
+          if (requirePrimary && !src.isPrimary) { continue; }
+          if (src.matchType !== 'exact') { continue; }
+          return true;
+        }
+        return false;
+      }
+      function collectExactHeadwordMatches(dictionaryEntries, token, requirePrimary) {
+        const matches = [];
+        for (const dictionaryEntry of dictionaryEntries || []) {
+          const headwords = Array.isArray(dictionaryEntry?.headwords) ? dictionaryEntry.headwords : [];
+          for (let headwordIndex = 0; headwordIndex < headwords.length; headwordIndex += 1) {
+            const headword = headwords[headwordIndex];
+            if (!hasExactSource(headword, token, requirePrimary)) { continue; }
+            matches.push({ dictionaryEntry, headword, headwordIndex });
+          }
+        }
+        return matches;
+      }
+      function sameHeadword(match, preferredMatch) {
+        if (!match || !preferredMatch) {
+          return false;
+        }
+        if (match.headword?.term !== preferredMatch.headword?.term) {
+          return false;
+        }
+        const matchReading = typeof match.headword?.reading === 'string' ? match.headword.reading : '';
+        const preferredReading =
+          typeof preferredMatch.headword?.reading === 'string' ? preferredMatch.headword.reading : '';
+        if (!matchReading || !preferredReading) {
+          return true;
+        }
+        return matchReading === preferredReading;
+      }
+      function getBestFrequencyRankForMatches(matches, dictionaryPriorityByName, dictionaryFrequencyModeByName) {
+        let best = null;
+        for (const match of matches) {
+          const rank = getBestFrequencyRank(
+            match.dictionaryEntry,
+            match.headwordIndex,
+            dictionaryPriorityByName,
+            dictionaryFrequencyModeByName
+          );
+          if (rank === null) { continue; }
+          if (best === null || rank < best) {
+            best = rank;
+          }
+        }
+        return best;
+      }
+`;