fix(anki): lower audio normalization target to -23 LUFS (EBU R128)

changes/audio-normalization.md

@@ -0,0 +1,4 @@
+type: fixed
+area: mining
+
+- Normalized generated card audio by default during media extraction, with `ankiConnect.media.normalizeAudio` available to keep raw source loudness when needed.

changes/youtube-stream-edl-options.md

@@ -0,0 +1,4 @@
+type: fixed
+area: youtube
+
+- Fixed direct YouTube stream media extraction by parsing mpv EDL stream URLs with their byte-length guards, preventing trailing EDL segment options from corrupting signed googlevideo URLs and causing ffmpeg 403 errors.

config.example.jsonc

@@ -559,6 +559,7 @@
       "animatedMaxHeight": 0, // Maximum height for animated AVIF captures, in pixels. Set to 0 to preserve aspect ratio.
       "animatedCrf": 35, // Animated AVIF CRF quality target. Lower values produce larger, higher-quality files.
       "syncAnimatedImageToWordAudio": true, // For animated AVIF images, prepend a frozen first frame matching the existing word-audio duration so motion starts with sentence audio. Values: true | false
+      "normalizeAudio": true, // Normalize generated sentence audio loudness during media extraction. Values: true | false
       "audioPadding": 0, // Seconds of padding appended to both ends of generated sentence audio and animated AVIF clips.
       "fallbackDuration": 3, // Fallback clip duration in seconds when subtitle timing data is unavailable.
       "maxMediaDuration": 30 // Maximum allowed media clip duration in seconds.

docs-site/anki-integration.md

@@ -161,13 +161,14 @@ Audio is extracted from the video file using the subtitle's start and end timest
 "ankiConnect": {
   "media": {
     "generateAudio": true,
+    "normalizeAudio": true,      // normalize generated clip loudness
     "audioPadding": 0,           // optional seconds before and after subtitle timing
     "maxMediaDuration": 30       // cap total duration in seconds
   }
 }
 ```
 
-Output format: MP3 at 44100 Hz. If the video has multiple audio streams, SubMiner uses the active stream.
+Output format: MP3 at 44100 Hz. If the video has multiple audio streams, SubMiner uses the active stream. Generated sentence audio is loudness-normalized by default during extraction; set `normalizeAudio` to `false` to keep raw source loudness.
 
 The audio is uploaded to Anki's media folder and inserted as `[sound:audio_<timestamp>.mp3]`.
 
@@ -347,6 +348,7 @@ When you mine the same word multiple times, SubMiner can merge the cards instead
       "imageType": "static",
       "imageFormat": "jpg",
       "imageQuality": 92,
+      "normalizeAudio": true,
       "audioPadding": 0,
       "maxMediaDuration": 30,
     },

docs-site/configuration.md

@@ -951,6 +951,7 @@ Enable automatic Anki card creation and updates with media generation:
       "animatedMaxWidth": 640,
       "animatedMaxHeight": 0,
       "animatedCrf": 35,
+      "normalizeAudio": true,
       "audioPadding": 0,
       "fallbackDuration": 3,
       "maxMediaDuration": 30
@@ -1001,6 +1002,7 @@ This example is intentionally compact. The option table below documents availabl
 | `ankiConnect.ai.model`                            | string                                      | Optional model override for Anki AI translation/enrichment flows.                                                                                                                                                               |
 | `ankiConnect.ai.systemPrompt`                     | string                                      | Optional system prompt override for Anki AI translation/enrichment flows.                                                                                                                                                       |
 | `media.generateAudio`                             | `true`, `false`                             | Generate audio clips from video (default: `true`)                                                                                                                                                                               |
+| `media.normalizeAudio`                            | `true`, `false`                             | Normalize generated sentence-audio loudness during media extraction (default: `true`). Set to `false` to keep raw source loudness.                                                                                              |
 | `media.generateImage`                             | `true`, `false`                             | Generate image/animation screenshots (default: `true`)                                                                                                                                                                          |
 | `media.imageType`                                 | `"static"`, `"avif"`                        | Image type: static screenshot or animated AVIF (default: `"static"`)                                                                                                                                                            |
 | `media.imageFormat`                               | `"jpg"`, `"png"`, `"webp"`                  | Image format (default: `"jpg"`)                                                                                                                                                                                                 |

docs-site/public/config.example.jsonc

@@ -559,6 +559,7 @@
       "animatedMaxHeight": 0, // Maximum height for animated AVIF captures, in pixels. Set to 0 to preserve aspect ratio.
       "animatedCrf": 35, // Animated AVIF CRF quality target. Lower values produce larger, higher-quality files.
       "syncAnimatedImageToWordAudio": true, // For animated AVIF images, prepend a frozen first frame matching the existing word-audio duration so motion starts with sentence audio. Values: true | false
+      "normalizeAudio": true, // Normalize generated sentence audio loudness during media extraction. Values: true | false
       "audioPadding": 0, // Seconds of padding appended to both ends of generated sentence audio and animated AVIF clips.
       "fallbackDuration": 3, // Fallback clip duration in seconds when subtitle timing data is unavailable.
       "maxMediaDuration": 30 // Maximum allowed media clip duration in seconds.

release/release-notes.md

@@ -1,28 +0,0 @@
-## Highlights
-
-### Fixed
-
-- **YouTube Background Cache:** Fixed Windows background media cache startup for YouTube URLs opened directly in mpv.
-  - Resolved stream URLs are now tracked even when mpv still exposes the original YouTube playlist entry.
-  - Queued Anki media updates can append audio and images after the cache finishes instead of staying text-only.
-
-- **YouTube Subtitle Picker Notifications:** Manual subtitle picker requests now show immediate status while SubMiner probes tracks and opens the modal.
-  - Subtitle download progress is replaced with a transient success notification after tracks load.
-
-## What's Changed
-
-- feat(youtube): notify on manual picker open and show success after track load by @ksyasuda in #133
-- fix(youtube): recover source URL for background media cache on direct mpv open by @ksyasuda in #132
-
-## Installation
-
-See the README and docs/installation guide for full setup steps.
-
-## Assets
-
-- Linux: `SubMiner.AppImage`
-- macOS: `SubMiner-*.dmg` and `SubMiner-*.zip`
-- Windows: `SubMiner-*.exe` and `SubMiner-*-win.zip`
-- Optional extras: `subminer-assets.tar.gz` and the `subminer` launcher
-
-Note: the `subminer` wrapper script uses Bun (`#!/usr/bin/env bun`), so `bun` must be installed and on `PATH`.

src/anki-integration.test.ts

@@ -860,13 +860,18 @@ test('AnkiIntegration queues YouTube media updates against recovered source URLs
   assert.equal(storedMedia.length, 2);
 });
 
-test('AnkiIntegration does not use mpv stream indexes for ready cached YouTube audio', async () => {
-  const audioCalls: Array<{ path: string; audioStreamIndex?: number }> = [];
+test('AnkiIntegration passes audio normalization config for ready cached YouTube audio', async () => {
+  const audioCalls: Array<{
+    path: string;
+    audioStreamIndex?: number;
+    normalizeAudio?: boolean;
+  }> = [];
 
   const integration = new AnkiIntegration(
     {
       media: {
         audioPadding: 0,
+        normalizeAudio: false,
       },
     },
     {} as never,
@@ -896,13 +901,21 @@ test('AnkiIntegration does not use mpv stream indexes for ready cached YouTube a
         endTime: number,
         audioPadding?: number,
         audioStreamIndex?: number,
+        normalizeAudio?: boolean,
       ) => Promise<Buffer>;
     };
     generateAudio: () => Promise<Buffer | null>;
   };
   internals.mediaGenerator = {
-    generateAudio: async (path, _startTime, _endTime, _audioPadding, audioStreamIndex) => {
-      audioCalls.push({ path: path.path, audioStreamIndex });
+    generateAudio: async (
+      path,
+      _startTime,
+      _endTime,
+      _audioPadding,
+      audioStreamIndex,
+      normalizeAudio,
+    ) => {
+      audioCalls.push({ path: path.path, audioStreamIndex, normalizeAudio });
       return Buffer.from('audio');
     },
   };
@@ -913,6 +926,7 @@ test('AnkiIntegration does not use mpv stream indexes for ready cached YouTube a
     {
       path: '/tmp/subminer-youtube-media-cache/media.mkv',
       audioStreamIndex: undefined,
+      normalizeAudio: false,
     },
   ]);
 });

src/anki-integration.ts

@@ -348,6 +348,7 @@ export class AnkiIntegration {
             endTime,
             audioPadding,
             audioStreamIndex,
+            this.config.media?.normalizeAudio !== false,
           ),
         generateScreenshot: (videoPath, timestamp, options) =>
           this.mediaGenerator.generateScreenshot(videoPath, timestamp, options),
@@ -502,6 +503,7 @@ export class AnkiIntegration {
             endTime,
             audioPadding,
             audioStreamIndex,
+            this.config.media?.normalizeAudio !== false,
           ),
         generateScreenshot: (videoPath, timestamp, options) =>
           this.mediaGenerator.generateScreenshot(videoPath, timestamp, options),
@@ -996,6 +998,7 @@ export class AnkiIntegration {
       endTime,
       this.config.media?.audioPadding,
       resolveAudioStreamIndexForMediaGeneration(videoPath, this.mpvClient.currentAudioStreamIndex),
+      this.config.media?.normalizeAudio !== false,
     );
   }
 

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

@@ -2,6 +2,7 @@ import assert from 'node:assert/strict';
 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';
 
@@ -269,9 +270,11 @@ test('manual clipboard subtitle update uses resolved mpv stream URLs for remote
   const imagePaths: string[] = [];
   const recordMediaPath = (mediaInput: MediaInput): string =>
     typeof mediaInput === 'string' ? mediaInput : mediaInput.path;
+  const audioUrl = 'https://audio.example/videoplayback?mime=audio%2Fwebm';
+  const videoUrl = 'https://video.example/videoplayback?mime=video%2Fmp4';
   const edlSource = [
-    'edl://!new_stream;!no_clip;!no_chapters;%70%https://audio.example/videoplayback?mime=audio%2Fwebm',
-    '!new_stream;!no_clip;!no_chapters;%69%https://video.example/videoplayback?mime=video%2Fmp4',
+    `edl://!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(audioUrl)}`,
+    `!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(videoUrl)}`,
     '!global_tags,title=test',
   ].join(';');
 
@@ -354,8 +357,8 @@ test('manual clipboard subtitle update uses resolved mpv stream URLs for remote
 
   await service.updateLastAddedFromClipboard('一行目\n\n二行目');
 
-  assert.deepEqual(audioPaths, ['https://audio.example/videoplayback?mime=audio%2Fwebm']);
-  assert.deepEqual(imagePaths, ['https://video.example/videoplayback?mime=video%2Fmp4']);
+  assert.deepEqual(audioPaths, [audioUrl]);
+  assert.deepEqual(imagePaths, [videoUrl]);
   assert.equal(storedMedia.length, 2);
   assert.equal(updatedFields.length, 1);
   assert.equal(updatedFields[0]?.Sentence, '一行目 二行目');

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

@@ -2,6 +2,7 @@ import assert from 'node:assert/strict';
 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';
 
@@ -290,9 +291,11 @@ test('CardCreationService uses stream-open-filename for remote media generation'
   const imagePaths: string[] = [];
   const recordMediaPath = (mediaInput: MediaInput): string =>
     typeof mediaInput === 'string' ? mediaInput : mediaInput.path;
+  const audioUrl = 'https://audio.example/videoplayback?mime=audio%2Fwebm';
+  const videoUrl = 'https://video.example/videoplayback?mime=video%2Fmp4';
   const edlSource = [
-    'edl://!new_stream;!no_clip;!no_chapters;%70%https://audio.example/videoplayback?mime=audio%2Fwebm',
-    '!new_stream;!no_clip;!no_chapters;%69%https://video.example/videoplayback?mime=video%2Fmp4',
+    `edl://!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(audioUrl)}`,
+    `!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(videoUrl)}`,
     '!global_tags,title=test',
   ].join(';');
 
@@ -397,8 +400,8 @@ test('CardCreationService uses stream-open-filename for remote media generation'
   const created = await service.createSentenceCard('テスト', 0, 1);
 
   assert.equal(created, true);
-  assert.deepEqual(audioPaths, ['https://audio.example/videoplayback?mime=audio%2Fwebm']);
-  assert.deepEqual(imagePaths, ['https://video.example/videoplayback?mime=video%2Fmp4']);
+  assert.deepEqual(audioPaths, [audioUrl]);
+  assert.deepEqual(imagePaths, [videoUrl]);
 });
 
 test('CardCreationService does not use mpv stream indexes for ready cached YouTube media', async () => {

src/anki-integration/card-creation.ts

@@ -65,6 +65,7 @@ interface CardCreationMediaGenerator {
     endTime: number,
     audioPadding?: number,
     audioStreamIndex?: number,
+    normalizeAudio?: boolean,
   ): Promise<Buffer | null>;
   generateScreenshot(
     path: MediaInput,
@@ -842,6 +843,7 @@ export class CardCreationService {
         videoPath,
         mpvClient.currentAudioStreamIndex ?? undefined,
       ),
+      this.deps.getConfig().media?.normalizeAudio !== false,
     );
   }
 

src/anki-integration/media-source.test.ts

@@ -2,6 +2,7 @@ import assert from 'node:assert/strict';
 import test from 'node:test';
 
 import * as mediaSource from './media-source';
+import { toMpvEdlValue } from './mpv-edl-test-utils';
 
 const { resolveMediaGenerationInputPath } = mediaSource;
 
@@ -53,9 +54,11 @@ test('resolveMediaGenerationInputPath prefers stream-open-filename for remote me
 });
 
 test('resolveMediaGenerationInputPath unwraps mpv edl source for audio and video', async () => {
+  const audioUrl = 'https://audio.example/videoplayback?mime=audio%2Fwebm';
+  const videoUrl = 'https://video.example/videoplayback?mime=video%2Fmp4';
   const edlSource = [
-    'edl://!new_stream;!no_clip;!no_chapters;%70%https://audio.example/videoplayback?mime=audio%2Fwebm',
-    '!new_stream;!no_clip;!no_chapters;%69%https://video.example/videoplayback?mime=video%2Fmp4',
+    `edl://!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(audioUrl)}`,
+    `!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(videoUrl)}`,
     '!global_tags,title=test',
   ].join(';');
 
@@ -74,8 +77,52 @@ test('resolveMediaGenerationInputPath unwraps mpv edl source for audio and video
     'video',
   );
 
-  assert.equal(audioResult, 'https://audio.example/videoplayback?mime=audio%2Fwebm');
-  assert.equal(videoResult, 'https://video.example/videoplayback?mime=video%2Fmp4');
+  assert.equal(audioResult, audioUrl);
+  assert.equal(videoResult, videoUrl);
+});
+
+test('resolveMediaGenerationInputPath strips mpv edl segment options from unwrapped streams', async () => {
+  const audioUrl = 'https://audio.example/videoplayback?mime=audio%2Fwebm';
+  const signedVideoUrl =
+    'https://rr1---sn.example.googlevideo.com/videoplayback?mime=video%2Fmp4&mn=sn-a,sn-b&lsig=abc%3D';
+  const edlSource = [
+    `edl://!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(audioUrl)}`,
+    `!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(signedVideoUrl)},title=clip,length=73,timestamps=chapters`,
+    '!global_tags,title=test',
+  ].join(';');
+
+  const result = await resolveMediaGenerationInputPath(
+    {
+      currentVideoPath: 'https://www.youtube.com/watch?v=abc123',
+      requestProperty: async () => edlSource,
+    },
+    'video',
+  );
+
+  assert.equal(result, signedVideoUrl);
+});
+
+test('resolveMediaGenerationInputPath ignores length-guarded URLs in mpv edl headers', async () => {
+  const initUrl = 'https://init.example/init.mp4';
+  const audioUrl = 'https://audio.example/stream';
+  const videoUrl = 'https://video.example/stream';
+  const edlSource = [
+    `edl://!mp4_dash,init=${toMpvEdlValue(initUrl)}`,
+    '!new_stream',
+    toMpvEdlValue(audioUrl),
+    '!new_stream',
+    toMpvEdlValue(videoUrl),
+  ].join(';');
+
+  const audioResult = await resolveMediaGenerationInputPath(
+    {
+      currentVideoPath: 'https://www.youtube.com/watch?v=abc123',
+      requestProperty: async () => edlSource,
+    },
+    'audio',
+  );
+
+  assert.equal(audioResult, audioUrl);
 });
 
 test('resolveMediaGenerationInputPath falls back to currentVideoPath when stream-open-filename fails', async () => {
@@ -97,9 +144,11 @@ test('resolveMediaGenerationInput returns single-stream metadata for mpv EDL URL
   ).resolveMediaGenerationInput;
   assert.equal(typeof resolver, 'function');
 
+  const audioUrl = 'https://audio.example/videoplayback?mime=audio%2Fwebm';
+  const videoUrl = 'https://video.example/videoplayback?mime=video%2Fmp4';
   const edlSource = [
-    'edl://!new_stream;!no_clip;!no_chapters;%70%https://audio.example/videoplayback?mime=audio%2Fwebm',
-    '!new_stream;!no_clip;!no_chapters;%69%https://video.example/videoplayback?mime=video%2Fmp4',
+    `edl://!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(audioUrl)}`,
+    `!new_stream;!no_clip;!no_chapters;${toMpvEdlValue(videoUrl)}`,
   ].join(';');
 
   const result = await resolver!(
@@ -117,7 +166,7 @@ test('resolveMediaGenerationInput returns single-stream metadata for mpv EDL URL
     'audio',
   );
 
-  assert.equal(result?.path, 'https://audio.example/videoplayback?mime=audio%2Fwebm');
+  assert.equal(result?.path, audioUrl);
   assert.equal(result?.singleResolvedStream, true);
   assert.equal(result?.inputOptions?.reconnect, true);
   assert.equal(result?.inputOptions?.userAgent, 'Mozilla/5.0');

src/anki-integration/media-source.ts

@@ -1,6 +1,7 @@
 import { isRemoteMediaPath } from '../jimaku/utils';
 import type { MediaInput, MediaInputOptions } from '../media-input';
 import type { MpvClient } from '../types/runtime';
+import { extractFileUrlsFromMpvEdlSource } from './mpv-edl';
 
 export type MediaGenerationKind = 'audio' | 'video';
 export type MediaGenerationInputSource =
@@ -73,9 +74,8 @@ function normalizeHeaderName(value: string): string | null {
 }
 
 function extractUrlsFromMpvEdlSource(source: string): string[] {
-  const matches = source.matchAll(/%\d+%(https?:\/\/.*?)(?=;!new_stream|;!global_tags|$)/gms);
-  return [...matches]
-    .map((match) => trimToNonEmptyString(match[1]))
+  return extractFileUrlsFromMpvEdlSource(source)
+    .map((value) => trimToNonEmptyString(value))
     .filter((value): value is string => value !== null);
 }
 

src/anki-integration/mpv-edl-test-utils.ts

@@ -0,0 +1,3 @@
+export function toMpvEdlValue(value: string): string {
+  return `%${Buffer.byteLength(value, 'utf8')}%${value}`;
+}

src/anki-integration/mpv-edl.test.ts

@@ -0,0 +1,37 @@
+import assert from 'node:assert/strict';
+import test from 'node:test';
+
+import { extractFileUrlsFromMpvEdlSource } from './mpv-edl';
+import { toMpvEdlValue } from './mpv-edl-test-utils';
+
+test('extractFileUrlsFromMpvEdlSource honors length-guarded file values', () => {
+  const url =
+    'https://rr1---sn.example.googlevideo.com/videoplayback?mime=video%2Fmp4&mn=sn-a,sn-b&lsig=abc%3D';
+  const source = `edl://!new_stream;${toMpvEdlValue(url)},title=clip,length=73`;
+
+  assert.deepEqual(extractFileUrlsFromMpvEdlSource(source), [url]);
+});
+
+test('extractFileUrlsFromMpvEdlSource reads file parameters', () => {
+  const initUrl = 'https://init.example/init.mp4';
+  const fileUrl = 'https://video.example/videoplayback?mime=video%2Fmp4';
+  const source = `edl://!mp4_dash,init=${toMpvEdlValue(initUrl)};file=${toMpvEdlValue(
+    fileUrl,
+  )},length=42`;
+
+  assert.deepEqual(extractFileUrlsFromMpvEdlSource(source), [fileUrl]);
+});
+
+test('extractFileUrlsFromMpvEdlSource aggregates file URLs across entries', () => {
+  const audioUrl = 'https://audio.example/videoplayback?mime=audio%2Fwebm';
+  const videoUrl = 'https://video.example/videoplayback?mime=video%2Fmp4';
+  const source = [
+    'edl://!new_stream',
+    toMpvEdlValue(audioUrl),
+    '!new_stream',
+    `file=${toMpvEdlValue(videoUrl)},length=50`,
+    '!global_tags,title=test',
+  ].join(';');
+
+  assert.deepEqual(extractFileUrlsFromMpvEdlSource(source), [audioUrl, videoUrl]);
+});

src/anki-integration/mpv-edl.ts

@@ -0,0 +1,195 @@
+const EDL_URI_PREFIX = 'edl://';
+
+const BYTE_COMMA = ','.charCodeAt(0);
+const BYTE_CR = '\r'.charCodeAt(0);
+const BYTE_EQUALS = '='.charCodeAt(0);
+const BYTE_EXCLAMATION = '!'.charCodeAt(0);
+const BYTE_LF = '\n'.charCodeAt(0);
+const BYTE_PERCENT = '%'.charCodeAt(0);
+const BYTE_SEMICOLON = ';'.charCodeAt(0);
+
+function isDigitByte(value: number | undefined): value is number {
+  return value !== undefined && value >= 48 && value <= 57;
+}
+
+function isEntrySeparator(value: number | undefined): boolean {
+  return value === BYTE_SEMICOLON || value === BYTE_LF || value === BYTE_CR;
+}
+
+function isParamSeparator(value: number | undefined): boolean {
+  return value === BYTE_COMMA || isEntrySeparator(value);
+}
+
+function decodeBytes(buffer: Buffer, start: number, end: number): string {
+  return buffer.subarray(start, end).toString('utf8');
+}
+
+function isHttpUrl(value: string): boolean {
+  return /^https?:\/\//i.test(value);
+}
+
+function toEdlDataBuffer(source: string): Buffer {
+  const data = source.startsWith(EDL_URI_PREFIX) ? source.slice(EDL_URI_PREFIX.length) : source;
+  return Buffer.from(data, 'utf8');
+}
+
+function parseLengthGuardedValue(
+  buffer: Buffer,
+  position: number,
+): { value: string; end: number } | null {
+  if (buffer[position] !== BYTE_PERCENT) {
+    return null;
+  }
+
+  let cursor = position + 1;
+  if (!isDigitByte(buffer[cursor])) {
+    return null;
+  }
+
+  let byteLength = 0;
+  while (true) {
+    const digit = buffer[cursor];
+    if (!isDigitByte(digit)) {
+      break;
+    }
+    byteLength = byteLength * 10 + (digit - 48);
+    cursor += 1;
+  }
+
+  if (buffer[cursor] !== BYTE_PERCENT) {
+    return null;
+  }
+
+  const valueStart = cursor + 1;
+  const valueEnd = valueStart + byteLength;
+  if (valueEnd > buffer.length) {
+    return null;
+  }
+
+  return {
+    value: decodeBytes(buffer, valueStart, valueEnd),
+    end: valueEnd,
+  };
+}
+
+function skipEntrySeparators(buffer: Buffer, position: number): number {
+  let cursor = position;
+  while (cursor < buffer.length && isEntrySeparator(buffer[cursor])) {
+    cursor += 1;
+  }
+  return cursor;
+}
+
+function skipEntry(buffer: Buffer, position: number): number {
+  let cursor = position;
+  while (cursor < buffer.length) {
+    const guardedValue = parseLengthGuardedValue(buffer, cursor);
+    if (guardedValue) {
+      cursor = guardedValue.end;
+      continue;
+    }
+    if (isEntrySeparator(buffer[cursor])) {
+      break;
+    }
+    cursor += 1;
+  }
+  return cursor;
+}
+
+function parseRawValue(buffer: Buffer, position: number): { value: string; end: number } {
+  let cursor = position;
+  while (
+    cursor < buffer.length &&
+    !isParamSeparator(buffer[cursor]) &&
+    buffer[cursor] !== BYTE_EXCLAMATION
+  ) {
+    cursor += 1;
+  }
+  return {
+    value: decodeBytes(buffer, position, cursor),
+    end: cursor,
+  };
+}
+
+function parseParamValue(buffer: Buffer, position: number): { value: string; end: number } {
+  return parseLengthGuardedValue(buffer, position) ?? parseRawValue(buffer, position);
+}
+
+function parseOptionalParamName(
+  buffer: Buffer,
+  position: number,
+): { name: string | null; valueStart: number } {
+  let cursor = position;
+  while (
+    cursor < buffer.length &&
+    !isParamSeparator(buffer[cursor]) &&
+    buffer[cursor] !== BYTE_PERCENT &&
+    buffer[cursor] !== BYTE_EXCLAMATION
+  ) {
+    if (buffer[cursor] === BYTE_EQUALS) {
+      return {
+        name: decodeBytes(buffer, position, cursor),
+        valueStart: cursor + 1,
+      };
+    }
+    cursor += 1;
+  }
+
+  return { name: null, valueStart: position };
+}
+
+function parseSegmentEntry(buffer: Buffer, position: number): { urls: string[]; end: number } {
+  const urls: string[] = [];
+  let cursor = position;
+  let unnamedParamIndex = 0;
+
+  while (cursor < buffer.length && !isEntrySeparator(buffer[cursor])) {
+    const { name, valueStart } = parseOptionalParamName(buffer, cursor);
+    const value = parseParamValue(buffer, valueStart);
+    const lowerName = name?.toLowerCase() ?? null;
+    const isFileParam = lowerName === 'file' || (lowerName === null && unnamedParamIndex === 0);
+
+    if (isFileParam && isHttpUrl(value.value)) {
+      urls.push(value.value);
+    }
+
+    if (lowerName === null) {
+      unnamedParamIndex += 1;
+    }
+
+    cursor = value.end;
+    if (buffer[cursor] === BYTE_COMMA) {
+      cursor += 1;
+      continue;
+    }
+    if (!isEntrySeparator(buffer[cursor])) {
+      cursor = skipEntry(buffer, cursor);
+    }
+  }
+
+  return { urls, end: cursor };
+}
+
+export function extractFileUrlsFromMpvEdlSource(source: string): string[] {
+  const buffer = toEdlDataBuffer(source);
+  const urls: string[] = [];
+  let cursor = 0;
+
+  while (cursor < buffer.length) {
+    cursor = skipEntrySeparators(buffer, cursor);
+    if (cursor >= buffer.length) {
+      break;
+    }
+
+    if (buffer[cursor] === BYTE_EXCLAMATION) {
+      cursor = skipEntry(buffer, cursor);
+      continue;
+    }
+
+    const segment = parseSegmentEntry(buffer, cursor);
+    urls.push(...segment.urls);
+    cursor = segment.end;
+  }
+
+  return urls;
+}

src/anki-integration/pending-youtube-media-queue.ts

@@ -272,6 +272,7 @@ export class PendingYoutubeMediaQueue {
           job.endTime,
           config.media?.audioPadding,
           undefined,
+          config.media?.normalizeAudio !== false,
         );
         if (audioBuffer) {
           await this.deps.client.storeMediaFile(audioFilename, audioBuffer);

src/config/config.test.ts

@@ -92,6 +92,7 @@ test('loads defaults when config is missing', () => {
     model: '',
     systemPrompt: '',
   });
+  assert.equal(config.ankiConnect.media.normalizeAudio, true);
   assert.equal(config.startupWarmups.lowPowerMode, false);
   assert.equal(config.startupWarmups.mecab, true);
   assert.equal(config.startupWarmups.yomitanExtension, true);

src/config/definitions/defaults-integrations.ts

@@ -51,6 +51,7 @@ export const INTEGRATIONS_DEFAULT_CONFIG: Pick<
       animatedMaxHeight: 0,
       animatedCrf: 35,
       syncAnimatedImageToWordAudio: true,
+      normalizeAudio: true,
       audioPadding: 0,
       fallbackDuration: 3.0,
       maxMediaDuration: 30,

src/config/definitions/domain-registry.test.ts

@@ -110,6 +110,7 @@ test('config option registry includes critical paths and has unique entries', ()
     'subtitleStyle.autoPauseVideoOnYomitanPopup',
     'ankiConnect.enabled',
     'subtitleStyle.nameMatchEnabled',
+    'ankiConnect.media.normalizeAudio',
     'anilist.characterDictionary.collapsibleSections.description',
     'mpv.executablePath',
     'mpv.launchMode',

src/config/definitions/options-integrations.ts

@@ -181,6 +181,12 @@ export function buildIntegrationConfigOptionRegistry(
       defaultValue: defaultConfig.ankiConnect.media.generateAudio,
       description: 'Generate sentence audio for mined cards.',
     },
+    {
+      path: 'ankiConnect.media.normalizeAudio',
+      kind: 'boolean',
+      defaultValue: defaultConfig.ankiConnect.media.normalizeAudio,
+      description: 'Normalize generated sentence audio loudness during media extraction.',
+    },
     {
       path: 'ankiConnect.media.generateImage',
       kind: 'boolean',

src/core/services/stats-server.ts

@@ -1206,6 +1206,7 @@ export function createStatsApp(
     const mediaGen = options?.createMediaGenerator?.() ?? new MediaGenerator();
 
     const audioPadding = ankiConfig.media?.audioPadding ?? 0;
+    const normalizeAudio = ankiConfig.media?.normalizeAudio !== false;
     const maxMediaDuration = ankiConfig.media?.maxMediaDuration ?? 30;
 
     const startSec = startMs / 1000;
@@ -1228,7 +1229,14 @@ export function createStatsApp(
 
     const audioPromise = generateAudio
       ? timeMiningPhase(mode, 'generateAudio', () =>
-          mediaGen.generateAudio(sourcePath, startSec, clampedEndSec, audioPadding),
+          mediaGen.generateAudio(
+            sourcePath,
+            startSec,
+            clampedEndSec,
+            audioPadding,
+            null,
+            normalizeAudio,
+          ),
         )
       : Promise.resolve(null);
 

src/media-generator.test.ts

@@ -163,6 +163,24 @@ test('generateAudio defaults to unpadded sentence timing', async () => {
   });
 });
 
+test('generateAudio normalizes sentence audio by default', async () => {
+  await withStubbedFfmpeg(async (generator, argsPath) => {
+    await generator.generateAudio('/video.mp4', 10, 12);
+
+    const args = readFfmpegArgs(argsPath);
+    assert.equal(args[args.indexOf('-af') + 1], 'loudnorm=I=-23:TP=-2:LRA=11');
+  });
+});
+
+test('generateAudio can preserve raw sentence audio loudness', async () => {
+  await withStubbedFfmpeg(async (generator, argsPath) => {
+    await generator.generateAudio('/video.mp4', 10, 12, 0, null, false);
+
+    const args = readFfmpegArgs(argsPath);
+    assert.equal(args.includes('-af'), false);
+  });
+});
+
 test('generateAudio clips leading padding without adding it to trailing duration', async () => {
   await withStubbedFfmpeg(async (generator, argsPath) => {
     await generator.generateAudio('/video.mp4', 0.2, 1.2, 0.5);

src/media-generator.ts

@@ -24,6 +24,7 @@ import { createLogger } from './logger';
 import { normalizeMediaInput, type MediaInput } from './media-input';
 
 const log = createLogger('media');
+const AUDIO_NORMALIZATION_FILTER = 'loudnorm=I=-23:TP=-2:LRA=11';
 
 export type { MediaInput, MediaInputOptions } from './media-input';
 
@@ -264,6 +265,7 @@ export class MediaGenerator {
     endTime: number,
     padding: number = 0,
     audioStreamIndex: number | null = null,
+    normalizeAudio = true,
   ): Promise<Buffer> {
     const safePadding = Number.isFinite(padding) ? Math.max(0, padding) : 0;
     const start = Math.max(0, startTime - safePadding);
@@ -293,7 +295,11 @@ export class MediaGenerator {
         args.push('-map', `0:${audioStreamIndex}`);
       }
 
-      args.push('-vn', '-acodec', 'libmp3lame', '-q:a', '2', '-ar', '44100', '-y', outputPath);
+      args.push('-vn');
+      if (normalizeAudio) {
+        args.push('-af', AUDIO_NORMALIZATION_FILTER);
+      }
+      args.push('-acodec', 'libmp3lame', '-q:a', '2', '-ar', '44100', '-y', outputPath);
 
       this.logMediaDebug(
         `audio start ${inputDescription} start=${start} duration=${duration} padding=${safePadding}`,

src/types/anki.ts

@@ -74,6 +74,7 @@ export interface AnkiConnectConfig {
     animatedMaxHeight?: number;
     animatedCrf?: number;
     syncAnimatedImageToWordAudio?: boolean;
+    normalizeAudio?: boolean;
     audioPadding?: number;
     fallbackDuration?: number;
     maxMediaDuration?: number;

src/types/config.ts

@@ -235,6 +235,7 @@ export interface ResolvedConfig {
       animatedMaxHeight?: number;
       animatedCrf: number;
       syncAnimatedImageToWordAudio: boolean;
+      normalizeAudio: boolean;
       audioPadding: number;
       fallbackDuration: number;
       maxMediaDuration: number;