The network-mount skip made SMB/NFS libraries fall back to live mpv text for
any release shipping subtitles only inside the container, losing karaoke
reconstruction, sidebar cues, and mining. The starvation it guarded against
had a different cause, and measured extraction runs at wire speed (~10s/GB
on gigabit) once per episode. Skip extraction only for true remote URLs,
which have no on-disk container to demux, and raise the extraction timeout
to cover large Bluray remuxes read over the network.
When embedded-track extraction is skipped (network-mounted media), live mpv
text during per-glyph typeset karaoke is a wall of simultaneous one-glyph
lines plus the syllable being typed. No parsed cues exist to substitute, so
the wall reached both overlays and recording verbatim. Detect bursts of many
single-glyph lines in the live fallback paths and drop them with their short
syllable companions, keeping concurrent dialogue lines.
Generated lyric effects can overlay each syllable with animated single letters
rendered through \fn in a symbol font, where ordinary letters draw as sparkles.
Reading them as text corrupted reconstructed lines ("sotto mimi ni ateru to a z
x") and leaked junk cues ("hlk"). A font a style group uses only for scattered
animated single glyphs now marks those events as decoration: they stay out of
fragment reconstruction and are suppressed alongside the line they overlay.
Per-glyph word gaps measured across a wide glyph ("waves|within" over s/w)
normalize to nearly a common advance, so the ratio test missed them. A word
space adds a roughly constant extra distance regardless of neighbor widths, so
glyph runs with enough gap samples also split when the advance exceeds the
width-predicted advance by a material fraction of the line's common unit.
Capital-to-lowercase pairs and short sample counts are excluded; both guards
are pinned by corpus-derived regression tests.
Across the 145-file library corpus this removes every scattered-letter
malformation and recovers 30+ missing word spaces with no other output change.
- Only classify tall positioned ASS blocks as fragment grids when they read like tiling (sign walls, re-shown countdown frames, scattered single glyphs, or table columns), keeping CC-style dialogue blocks and wrapped lyric rows publishable
- Widen Latin word-gap heuristics for per-glyph typesetting runs and short capitalized words so proportional-font variation and two-letter words no longer get split
- Add changelog fragment documenting the stats database busy-timeout fix
- Skip control-debris stripping for primary/secondary live text when cues or source aren't ASS/SSA, so SRT lines that merely resemble ASS override tags survive
- Track active source's ASS-ness in the secondary controller and gate handleLiveText's cleanup on it
- Recover Latin/romaji word spaces encoded only by positioned `\pos`/`\move` fragment gaps in ASS karaoke
- Drop malformed ASS rotation-reset spacer lines and skip zero-duration metadata events
- Dedupe duplicate lines across multiline primary/secondary cues, including full-width variants
- Treat dense fragment-grid sign layouts as visual typesetting, not publishable text, during primary subtitle resolution
- Build the Swift window-tracking helper with an explicit `-target ...-apple-macos12.0` instead of letting swiftc stamp the build machine's OS, so the overlay attaches to mpv on older macOS (e.g. Ventura) instead of crashing at load
- Fold the standalone build-macos-helper.sh into prepare-build-assets.mjs and update docs/tests to match
- Only trim indentation before slicing ASS event fields so a trailing authored space at a fragment's end is kept, preventing words from being joined together when fragmented karaoke lines are reconstructed
- Add regression test covering event-boundary word spacing
- Reconstruct fragment-only karaoke while preserving canonical animated signs
- Reprocess live subtitles after initial playback and seek updates
- Add a development launcher for playable media files
- Stop deleting imm_daily_rollups/imm_monthly_rollups on unrelated schema version bumps; their source session/telemetry rows may already be pruned, so deleted buckets could not be rebuilt
- Run startup session-rollup maintenance before the lexical rollup backfill takes the write lock, so recovery no longer races playback writes
- Update the vocabulary summary totals changelog fragment to reflect that watch-time, activity, efficiency, and library charts are no longer cleared during rebuilds
sudacode
