feat(asr): [PKG-ASR-09] add explicit capture ownership and cancel stale restarts

packages/artplayer-plugin-asr/src/audio-graph.ts

@@ -13,7 +13,7 @@ export class AudioGraph {
   private direct = false
   private workletLoaded = false
 
-  constructor(private readonly video: CaptureVideo, private readonly sampleRate: number) {}
+  constructor(private readonly video: CaptureVideo, private readonly sampleRate: number, private readonly captureOnly = false) {}
 
   private assertOpen() {
     if (this.closed)
@@ -49,12 +49,16 @@ export class AudioGraph {
     }
     this.assertOpen()
     if (!this.source) {
-      try {
-        this.source = context.createMediaElementSource(this.video)
-        this.direct = true
+      if (!this.captureOnly) {
+        try {
+          this.source = context.createMediaElementSource(this.video)
+          this.direct = true
+        }
+        catch (error) {
+          console.warn('[artplayerPluginAsr] Direct connection failed:', error)
+        }
       }
-      catch (error) {
-        console.warn('[artplayerPluginAsr] Direct connection failed:', error)
+      if (!this.source) {
         const capture = this.video.captureStream || this.video.mozCaptureStream
         if (!capture)
           throw new Error('Could not establish audio source')

packages/artplayer-plugin-asr/src/capture.ts

@@ -34,8 +34,11 @@ export class Capture {
   restart() {
     const active = this.running || Boolean(this.starting)
     this.pause()
-    if (this.graph && !this.graph.ownsMediaConnection)
-      return this.stop().then(() => active ? this.start() : undefined)
+    if (this.graph && !this.graph.ownsMediaConnection) {
+      const stopping = this.stop()
+      const epoch = this.epoch
+      return stopping.then(() => active && !this.terminal && this.epoch === epoch ? this.start() : undefined)
+    }
     return active ? this.start() : Promise.resolve()
   }
 
@@ -54,7 +57,7 @@ export class Capture {
       const count = Math.floor(sampleRate * interval / 1000)
       if (!Number.isSafeInteger(count) || count < 1 || !Number.isFinite(interval) || interval <= 0)
         throw new Error('Audio chunk length must be positive and finite')
-      const graph = this.graph ??= new AudioGraph(this.video, sampleRate)
+      const graph = this.graph ??= new AudioGraph(this.video, sampleRate, this.options.captureOnly)
       await graph.prepare()
       if (!current()) {
         if (!this.running && this.graph === graph)

packages/artplayer-plugin-asr/src/index.ts

@@ -5,13 +5,16 @@ import style from './style.less?inline'
 import { createSubtitles } from './subtitles'
 
 export default function artplayerPluginAsr(option: AsrOptions = {}) {
-  const { length = 3, interval = 100, sampleRate = 16000, autoHideTimeout = 10000, onAudioChunk = () => null } = option
+  const { length = 3, interval = 100, sampleRate = 16000, autoHideTimeout = 10000, onAudioChunk = () => null, audioInput } = option
+  if (audioInput !== undefined && audioInput.type !== 'capture')
+    throw new TypeError('Unsupported ASR audio input')
+  const captureOnly = audioInput?.type === 'capture'
   return (art: Artplayer): AsrResult => {
     const layer = art.layers.add({ name: 'asr', html: '' })
     if (!layer)
       throw new Error('Could not create ASR subtitle layer')
     const subtitles = createSubtitles(layer, length, autoHideTimeout)
-    const capture = new Capture(art.video, { interval, sampleRate, onAudioChunk }, subtitles.append)
+    const capture = new Capture(art.video, { interval, sampleRate, onAudioChunk, captureOnly }, subtitles.append)
     const play = () => capture.start()
     const pause = () => capture.pause()
     const restart = () => capture.restart()

packages/artplayer-plugin-asr/src/types.ts

@@ -4,6 +4,7 @@ export interface AudioChunk {
 }
 
 export interface AsrOptions {
+  audioInput?: { type: 'capture' }
   length?: number
   interval?: number
   sampleRate?: number
@@ -24,6 +25,7 @@ export interface CaptureVideo extends HTMLVideoElement {
 }
 
 export interface CaptureOptions {
+  captureOnly?: boolean
   interval: number
   sampleRate: number
   onAudioChunk: (chunk: AudioChunk) => unknown