fix(subtitles): follow mpv sub-delay in mining, sidebar, and navigation (#302)

This commit is contained in:
sudacode authored and GitHub committed 2026-10-26 22:19:08 -07:00
1 parent d3a94536a8
commit cc9af79d6e
25 files changed
+395 -54

No files matched your search

+1
View File
@@ -55,6 +55,7 @@ export {
subtitleCueListSeekTime,
subtitleCueSeekTime,
} from './subtitle-cue-navigation';
export { shiftSubtitleRangeToVideoClock, toSubtitleClock, toVideoClock } from './subtitle-clock';
export { createFrequencyDictionaryLookup } from './frequency-dictionary';
export { createJlptVocabularyLookup } from './jlpt-vocab';
export {
+1
View File
@@ -89,6 +89,7 @@ function createSubtitleSidebarSnapshotFixture(): SubtitleSidebarSnapshot {
cuesRevision: 1,
cues: [],
currentTimeSec: null,
subDelay: 0,
currentSubtitle: { text: '', startTime: null, endTime: null },
config: {
enabled: false,
+19 -1
View File
@@ -77,6 +77,7 @@ test('mineSentenceCard handles missing integration and disconnected mpv', async
currentSubText: 'line',
currentSubStart: 1,
currentSubEnd: 2,
currentSubDelay: 0,
},
showMpvOsd: (text) => osd.push(text),
}),
@@ -109,6 +110,7 @@ test('mineSentenceCard creates sentence card from mpv subtitle state', async ()
currentSubText: 'subtitle line',
currentSubStart: 10,
currentSubEnd: 12,
currentSubDelay: 0,
currentSecondarySubText: 'secondary line',
},
showMpvOsd: () => {},
@@ -148,6 +150,7 @@ test('mineSentenceCard prefers a canonical primary subtitle snapshot', async ()
currentSubText: '今今今手手手',
currentSubStart: 11.4,
currentSubEnd: 11.8,
currentSubDelay: 0,
currentSecondarySubText: 'English subtitle',
},
primarySubtitle: {
@@ -187,6 +190,7 @@ test('mineSentenceCard uses normalized secondary subtitle state instead of raw m
currentSubText: '日本語字幕',
currentSubStart: 10,
currentSubEnd: 12,
currentSubDelay: 0,
currentSecondarySubText: 'Your\nmosaic',
requestProperty: async () => {
requestedRawSecondaryText = true;
@@ -218,6 +222,7 @@ test('mineSentenceCard omits normalized secondary text that matches the primary
currentSubText: '日本語字幕',
currentSubStart: 10,
currentSubEnd: 12,
currentSubDelay: 0,
currentSecondarySubText: '日本語字幕',
},
showMpvOsd: () => {},
@@ -396,7 +401,7 @@ test('handleMineSentenceDigit joins per-entry secondary subtitles when available
test('captureLiveSubtitleMiningContext snapshots the current line and timings', () => {
const context = captureLiveSubtitleMiningContext(
{ currentSubText: ' 食べる ', currentSubStart: 12.5, currentSubEnd: 15.25 },
{ currentSubText: ' 食べる ', currentSubStart: 12.5, currentSubEnd: 15.25, currentSubDelay: 0 },
() => 1234,
);
@@ -409,6 +414,16 @@ test('captureLiveSubtitleMiningContext snapshots the current line and timings',
});
});
test('captureLiveSubtitleMiningContext moves sub-delay retimed lines onto the video clock', () => {
const context = captureLiveSubtitleMiningContext(
{ currentSubText: '食べる', currentSubStart: 5, currentSubEnd: 8, currentSubDelay: -1.5 },
() => 1234,
);
assert.equal(context?.startTime, 3.5);
assert.equal(context?.endTime, 6.5);
});
test('captureLiveSubtitleMiningContext rejects missing client, empty text and bad timings', () => {
assert.equal(captureLiveSubtitleMiningContext(null), null);
assert.equal(
@@ -416,6 +431,7 @@ test('captureLiveSubtitleMiningContext rejects missing client, empty text and ba
currentSubText: ' ',
currentSubStart: 1,
currentSubEnd: 2,
currentSubDelay: 0,
}),
null,
);
@@ -424,6 +440,7 @@ test('captureLiveSubtitleMiningContext rejects missing client, empty text and ba
currentSubText: 'line',
currentSubStart: 5,
currentSubEnd: 5,
currentSubDelay: 0,
}),
null,
);
@@ -432,6 +449,7 @@ test('captureLiveSubtitleMiningContext rejects missing client, empty text and ba
currentSubText: 'line',
currentSubStart: NaN,
currentSubEnd: 2,
currentSubDelay: 0,
}),
null,
);
+19 -4
View File
@@ -1,5 +1,6 @@
import type { SubtitleTimingBlock } from '../../subtitle-timing-tracker';
import type { SubtitleMiningContext } from '../../types/subtitle';
import { shiftSubtitleRangeToVideoClock } from './subtitle-clock';
interface SubtitleTimingTrackerLike {
getRecentBlocks: (count: number) => string[];
@@ -25,6 +26,7 @@ interface MpvClientLike {
currentSubText: string;
currentSubStart: number;
currentSubEnd: number;
currentSubDelay: number;
currentSecondarySubText?: string;
requestProperty?: (name: string) => Promise<unknown>;
}
@@ -62,7 +64,10 @@ export function handleMultiCopyDigit(
* line is on screen by then; callers capture here at mining time instead.
*/
export function captureLiveSubtitleMiningContext(
client: Pick<MpvClientLike, 'currentSubText' | 'currentSubStart' | 'currentSubEnd'> | null,
client: Pick<
MpvClientLike,
'currentSubText' | 'currentSubStart' | 'currentSubEnd' | 'currentSubDelay'
> | null,
now: () => number = Date.now,
): SubtitleMiningContext | null {
if (!client) {
@@ -73,7 +78,10 @@ export function captureLiveSubtitleMiningContext(
if (!text || !Number.isFinite(startTime) || !Number.isFinite(endTime) || endTime <= startTime) {
return null;
}
return { source: 'overlay', text, startTime, endTime, capturedAtMs: now() };
return shiftSubtitleRangeToVideoClock<SubtitleMiningContext>(
{ source: 'overlay', text, startTime, endTime, capturedAtMs: now() },
client.currentSubDelay,
);
}
export function copyCurrentSubtitle(deps: {
@@ -167,6 +175,7 @@ export async function markLastCardAsAudioCard(deps: {
export async function mineSentenceCard(deps: {
ankiIntegration: AnkiIntegrationLike | null;
mpvClient: MpvClientLike | null;
/** Already on the video clock, like every `SubtitleMiningContext`. */
primarySubtitle?: Pick<SubtitleMiningContext, 'text' | 'startTime' | 'endTime'>;
showMpvOsd: (text: string) => void;
}): Promise<boolean> {
@@ -185,10 +194,16 @@ export async function mineSentenceCard(deps: {
}
const secondarySubText = getCurrentSecondarySubTextForSentenceCard(mpvClient, primaryText);
const range =
deps.primarySubtitle ??
shiftSubtitleRangeToVideoClock(
{ startTime: mpvClient.currentSubStart, endTime: mpvClient.currentSubEnd },
mpvClient.currentSubDelay,
);
return await anki.createSentenceCard(
primaryText,
deps.primarySubtitle?.startTime ?? mpvClient.currentSubStart,
deps.primarySubtitle?.endTime ?? mpvClient.currentSubEnd,
range.startTime,
range.endTime,
secondarySubText,
);
}
+1
View File
@@ -38,6 +38,7 @@ const MPV_SUBTITLE_PROPERTY_OBSERVATIONS: string[] = [
'path',
'sub-start',
'sub-end',
'sub-delay',
'time-pos',
'secondary-sub-text',
'aid',
+41
View File
@@ -73,6 +73,8 @@ function createDeps(overrides: Partial<MpvProtocolHandleMessageDeps> = {}): {
getCurrentSubStart: () => 0,
setCurrentSubEnd: () => {},
getCurrentSubEnd: () => 0,
setCurrentSubDelay: () => {},
getCurrentSubDelay: () => 0,
emitMediaPathChange: (payload) => {
state.mediaPath = payload.path;
},
@@ -131,6 +133,22 @@ test('dispatchMpvProtocolMessage emits subtitle text on property change', async
assert.deepEqual(state.events, [{ text: '字幕', isOverlayVisible: false }]);
});
test('dispatchMpvProtocolMessage tracks sub-delay retiming', async () => {
const delays: number[] = [];
const { deps } = createDeps({ setCurrentSubDelay: (value) => delays.push(value) });
await dispatchMpvProtocolMessage(
{ event: 'property-change', name: 'sub-delay', data: 1.5 },
deps,
);
await dispatchMpvProtocolMessage(
{ event: 'property-change', name: 'sub-delay', data: null },
deps,
);
assert.deepEqual(delays, [1.5, 0]);
});
test('dispatchMpvProtocolMessage widens halfwidth katakana in subtitle text', async () => {
const { deps, state } = createDeps();
@@ -379,6 +397,29 @@ test('dispatchMpvProtocolMessage pauses on sub-end when pendingPauseAtSubEnd is
});
});
test('dispatchMpvProtocolMessage pauses at the sub-delay retimed sub-end', async () => {
let pauseAtTime: number | null = null;
let subDelay = 1.5;
const { deps } = createDeps({
getPendingPauseAtSubEnd: () => true,
getCurrentSubDelay: () => subDelay,
setCurrentSubDelay: (next) => {
subDelay = next;
},
getPauseAtTime: () => pauseAtTime,
setPauseAtTime: (next) => {
pauseAtTime = next;
},
});
await dispatchMpvProtocolMessage({ event: 'property-change', name: 'sub-end', data: 42 }, deps);
assert.equal(pauseAtTime, 43.5);
// Retiming before the replay reaches its end moves the pause with the line.
await dispatchMpvProtocolMessage({ event: 'property-change', name: 'sub-delay', data: 2 }, deps);
assert.equal(pauseAtTime, 44);
});
test('dispatchMpvProtocolMessage updates current time before emitting time-pos change', async () => {
const calls: string[] = [];
let currentTimePos = 0;
+15 -1
View File
@@ -1,6 +1,7 @@
import { MpvSubtitleRenderMetrics } from '../../types';
import { sanitizeMediaTitle } from '../../shared/media-identity';
import { widenHalfwidthKana } from '../utils/halfwidth-kana';
import { toVideoClock } from './subtitle-clock';
export type MpvMessage = {
event?: string;
@@ -64,6 +65,8 @@ export interface MpvProtocolHandleMessageDeps {
getCurrentSubStart: () => number;
setCurrentSubEnd: (value: number) => void;
getCurrentSubEnd: () => number;
setCurrentSubDelay: (value: number) => void;
getCurrentSubDelay: () => number;
emitMediaPathChange: (payload: { path: string }) => void;
emitMediaTitleChange: (payload: { title: string | null }) => void;
emitTimePosChange: (payload: { time: number }) => void;
@@ -263,7 +266,8 @@ export async function dispatchMpvProtocolMessage(
const subEnd = (msg.data as number) || 0;
deps.setCurrentSubEnd(subEnd);
if (deps.getPendingPauseAtSubEnd() && subEnd > 0) {
deps.setPauseAtTime(subEnd);
// `sub-end` is on the subtitle clock; the pause check compares it with `time-pos`.
deps.setPauseAtTime(toVideoClock(subEnd, deps.getCurrentSubDelay()));
deps.setPendingPauseAtSubEnd(false);
deps.sendCommand({ command: ['set_property', 'pause', false] });
}
@@ -272,6 +276,16 @@ export async function dispatchMpvProtocolMessage(
start: deps.getCurrentSubStart(),
end: deps.getCurrentSubEnd(),
});
} else if (msg.name === 'sub-delay') {
const parsed = typeof msg.data === 'number' ? msg.data : Number(msg.data);
const delay = Number.isFinite(parsed) ? parsed : 0;
// A scheduled replay pause sits at the line's end on the video clock; retiming
// mid-replay moves that end, so the pause moves with it.
const pauseAtTime = deps.getPauseAtTime();
if (pauseAtTime !== null) {
deps.setPauseAtTime(pauseAtTime + delay - deps.getCurrentSubDelay());
}
deps.setCurrentSubDelay(delay);
} else if (msg.name === 'secondary-sub-text') {
const nextSubText = (msg.data as string) || '';
deps.setCurrentSecondarySubText(nextSubText);
+6
View File
@@ -160,6 +160,8 @@ export class MpvIpcClient implements MpvClient {
public currentTimePos = 0;
public currentSubStart = 0;
public currentSubEnd = 0;
/** mpv `sub-delay`: seconds each primary line plays after its `sub-start`/`sub-end`. */
public currentSubDelay = 0;
public currentSubText = '';
public currentSecondarySubText = '';
public currentAudioStreamIndex: number | null = null;
@@ -459,6 +461,10 @@ export class MpvIpcClient implements MpvClient {
this.currentSubEnd = value;
},
getCurrentSubEnd: () => this.currentSubEnd,
setCurrentSubDelay: (value: number) => {
this.currentSubDelay = value;
},
getCurrentSubDelay: () => this.currentSubDelay,
emitMediaPathChange: (payload) => {
this.emit('media-path-change', payload);
},
+32
View File
@@ -0,0 +1,32 @@
/**
* mpv's `sub-start`/`sub-end` and parsed cues carry the subtitle file's own timestamps
* (the subtitle clock), but mpv plays each line `sub-delay` seconds later (z/x retiming,
* `sub-step`). Playback position, seeks, and media cuts are on the video clock, so any
* value crossing between the two converts here.
*/
function finiteDelay(subDelay: number): number {
return Number.isFinite(subDelay) ? subDelay : 0;
}
/** Playback position (`time-pos`) on the subtitle clock, comparable with cue times. */
export function toSubtitleClock(videoTimeSec: number, subDelay: number): number {
return videoTimeSec - finiteDelay(subDelay);
}
/** Subtitle-clock seconds on the video clock, for seeks. */
export function toVideoClock(subtitleTimeSec: number, subDelay: number): number {
return subtitleTimeSec + finiteDelay(subDelay);
}
/** A subtitle-clock range on the video clock, for mining and media generation. */
export function shiftSubtitleRangeToVideoClock<T extends { startTime: number; endTime: number }>(
range: T,
subDelay: number,
): T {
const delay = finiteDelay(subDelay);
if (delay === 0) {
return range;
}
return { ...range, startTime: range.startTime + delay, endTime: range.endTime + delay };
}
@@ -26,7 +26,7 @@ test('next subtitle navigation skips generated ASS events and seeks to the next
},
];
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 10.2), [
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 10.2, 0), [
'seek',
13.08,
'absolute+exact',
@@ -41,7 +41,7 @@ test('next subtitle navigation treats simultaneous sanitized cues as one line bo
{ startTime: 13.02, endTime: 16, text: 'Next English' },
];
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 10.1), [
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 10.1, 0), [
'seek',
13.08,
'absolute+exact',
@@ -55,7 +55,7 @@ test('next subtitle navigation advances past the latest overlapping lyric', () =
{ startTime: 16, endTime: 19, text: 'next lyric' },
];
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 13.2), [
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 13.2, 0), [
'seek',
16.08,
'absolute+exact',
@@ -68,7 +68,7 @@ test('previous subtitle navigation leaves the current cue and seeks to the prior
{ startTime: 13, endTime: 16, text: 'current line' },
];
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', -1], cues, 14.5), [
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', -1], cues, 14.5, 0), [
'seek',
10.08,
'absolute+exact',
@@ -78,8 +78,22 @@ test('previous subtitle navigation leaves the current cue and seeks to the prior
test('subtitle navigation falls back when no sanitized destination exists', () => {
const cues = [{ startTime: 10, endTime: 13, text: 'only line' }];
assert.equal(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 10.2), null);
assert.equal(resolveSanitizedSubtitleSeekCommand(['seek', 5], cues, 10.2), null);
assert.equal(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 10.2, 0), null);
assert.equal(resolveSanitizedSubtitleSeekCommand(['seek', 5], cues, 10.2, 0), null);
});
test('subtitle cue navigation finds and seeks cues on the sub-delay retimed timeline', () => {
const cues = [
{ startTime: 10, endTime: 13, text: 'one' },
{ startTime: 13, endTime: 16, text: 'two' },
];
// At 14.5 with a 2s delay, mpv is showing "one" (12-15 on the video clock).
assert.deepEqual(resolveSanitizedSubtitleSeekCommand(['sub-seek', 1], cues, 14.5, 2), [
'seek',
15.08,
'absolute+exact',
]);
});
test('sidebar cue seeks share the boundary-safe sanitized cue timestamp', () => {
+5 -2
View File
@@ -1,3 +1,4 @@
import { toSubtitleClock, toVideoClock } from './subtitle-clock';
import type { SubtitleCue } from './subtitle-cue-parser';
const CUE_START_GROUP_TOLERANCE_SECONDS = 0.05;
@@ -80,8 +81,10 @@ export function subtitleCueListSeekTime(
export function resolveSanitizedSubtitleSeekCommand(
command: readonly (string | number)[],
cues: readonly SubtitleCue[],
currentTimeSec: number,
videoTimeSec: number,
subDelay: number,
): (string | number)[] | null {
const currentTimeSec = toSubtitleClock(videoTimeSec, subDelay);
if (
command.length < 2 ||
command[0] !== 'sub-seek' ||
@@ -124,5 +127,5 @@ export function resolveSanitizedSubtitleSeekCommand(
if (!destination) {
return null;
}
return ['seek', subtitleCueSeekTime(destination.cue), 'absolute+exact'];
return ['seek', toVideoClock(subtitleCueSeekTime(destination.cue), subDelay), 'absolute+exact'];
}