feat(dictation): stop desktop dictation after a minute of silence (#39131)

* feat(dictation): stop desktop dictation after a minute of silence

A toggle dictation left running keeps the mic open indefinitely. Stop it
a minute after the last speech, with a toast, like the phone's tap mode.
A held push-to-talk key is never stopped; its release ends the session.

Mobile's timer moves into src/shared so both apps use the same rule, and
the shared meter now records when it last heard speech.

* fix(dictation): gate the silence stop on hold mode, as the phone does

* refactor(dictation): let the desktop silence hook own the hold-mode rule, as mobile's does

* refactor(dictation): move the capture meter into its own hook

* refactor(dictation): initialise the capture meter ref without a render write

mobile/src/session/use-mobile-dictation-silence-stop.ts

@@ -1,9 +1,8 @@
 import { useEffect, useEffectEvent } from 'react'
+import { watchDictationSilence } from '../../../src/shared/dictation-silence-stop'
 import type { UseMobileDictationResult } from '../hooks/mobile-dictation-session-state'
 import { nativeChatDictationCopy } from './mobile-session-native-chat-dictation-copy'
 
-const DICTATION_SILENCE_STOP_MS = 60_000
-
 // A forgotten tap-mode mic keeps the screen awake; a finger holding the button is never forgotten.
 export function useMobileDictationSilenceStop(
   dictation: Pick<UseMobileDictationResult, 'isRecording' | 'meter'>,
@@ -22,19 +21,6 @@ export function useMobileDictationSilenceStop(
     if (!watching) {
       return
     }
-    // Wall clock, so a stalled audio stream still counts as silence.
-    const recordingAt = Date.now()
-    let timer: ReturnType<typeof setTimeout>
-    const check = () => {
-      const lastSpeechAt = Math.max(recordingAt, meter.lastSpeechAt())
-      const remainingMs = lastSpeechAt + DICTATION_SILENCE_STOP_MS - Date.now()
-      if (remainingMs > 0) {
-        timer = setTimeout(check, remainingMs)
-      } else {
-        stopForSilence()
-      }
-    }
-    check()
-    return () => clearTimeout(timer)
+    return watchDictationSilence(meter.lastSpeechAt, Date.now, () => stopForSilence())
   }, [meter, watching])
 }

src/renderer/src/components/dictation/DictationController.tsx

@@ -19,6 +19,7 @@ import { useHoldDictationGesture } from './use-hold-dictation-gesture'
 import { DICTATION_CONTROL_EVENT, type DictationControlAction } from './dictation-control-events'
 import { publishDictationMeter } from './dictation-meter-store'
 import { stopDictationAtAudioLimit } from './dictation-audio-limit-stop'
+import { useDictationSilenceStop } from './use-dictation-silence-stop'
 
 export function DictationController() {
   const dictationState = useAppStore((s) => s.dictationState)
@@ -33,7 +34,8 @@ export function DictationController() {
     finish: finishCapture,
     flushBufferedAudio,
     discardBufferedAudio,
-    getCapturedChunkCount
+    getCapturedChunkCount,
+    getLastSpeechAt
   } = useAudioCapture(publishDictationMeter)
 
   const dictationStateRef = useRef(dictationState)
@@ -319,6 +321,13 @@ export function DictationController() {
     return () => document.removeEventListener(DICTATION_CONTROL_EVENT, handleControl)
   }, [settings?.voice?.enabled, startDictation, stopDictation])
 
+  useDictationSilenceStop(
+    dictationState === 'listening',
+    settings?.voice?.dictationMode,
+    getLastSpeechAt,
+    () => void stopDictation()
+  )
+
   useHoldDictationGesture({
     dictationStateRef,
     keybindings,

src/renderer/src/components/dictation/use-dictation-silence-stop.ts

@@ -0,0 +1,34 @@
+import { useEffect, useEffectEvent } from 'react'
+import { toast } from 'sonner'
+import { translate } from '@/i18n/i18n'
+import { watchDictationSilence } from '../../../../shared/dictation-silence-stop'
+
+// A forgotten toggle mic stays open; a held push-to-talk key or button ends the session on release.
+export function useDictationSilenceStop(
+  listening: boolean,
+  dictationMode: string | undefined,
+  getLastSpeechAt: () => number,
+  stopDictation: () => void
+): void {
+  const watching = listening && dictationMode !== 'hold'
+  const stopForSilence = useEffectEvent(() => {
+    toast.message(
+      translate(
+        'auto.components.dictation.DictationController.stoppedAfterSilence',
+        'Stopped after a minute of silence'
+      )
+    )
+    stopDictation()
+  })
+
+  useEffect(() => {
+    if (!watching) {
+      return
+    }
+    return watchDictationSilence(
+      getLastSpeechAt,
+      () => performance.now(),
+      () => stopForSilence()
+    )
+  }, [watching, getLastSpeechAt])
+}

src/renderer/src/hooks/use-audio-capture-meter.ts

@@ -0,0 +1,49 @@
+import { useCallback, useRef } from 'react'
+import {
+  DEFAULT_DICTATION_METER,
+  analyzeDictationAudioChunk,
+  createDictationMeterAnalyzerState,
+  dictationMeterStatesEqual,
+  toPublicDictationMeterState,
+  type DictationMeterAnalyzerState,
+  type DictationMeterState
+} from '../../../shared/dictation-audio-meter'
+
+const METER_PUBLISH_INTERVAL_MS = 1000 / 15
+
+export type DictationMeterPublisher = (meter: DictationMeterState) => void
+
+export function useAudioCaptureMeter(publishMeter?: DictationMeterPublisher) {
+  const analyzerRef = useRef<DictationMeterAnalyzerState>(createDictationMeterAnalyzerState())
+  const publishedMeterRef = useRef(DEFAULT_DICTATION_METER)
+  const lastPublishedAtRef = useRef(Number.NEGATIVE_INFINITY)
+
+  const resetMeter = useCallback(() => {
+    analyzerRef.current = createDictationMeterAnalyzerState()
+    publishedMeterRef.current = DEFAULT_DICTATION_METER
+    lastPublishedAtRef.current = Number.NEGATIVE_INFINITY
+    publishMeter?.(DEFAULT_DICTATION_METER)
+  }, [publishMeter])
+
+  const measureChunk = useCallback(
+    (samples: Float32Array) => {
+      const now = performance.now()
+      analyzerRef.current = analyzeDictationAudioChunk(samples, now, analyzerRef.current)
+      if (document.hidden || now - lastPublishedAtRef.current < METER_PUBLISH_INTERVAL_MS) {
+        return
+      }
+      lastPublishedAtRef.current = now
+      const meter = toPublicDictationMeterState(analyzerRef.current)
+      if (!dictationMeterStatesEqual(publishedMeterRef.current, meter)) {
+        publishedMeterRef.current = meter
+        publishMeter?.(meter)
+      }
+    },
+    [publishMeter]
+  )
+
+  // On the performance.now() clock measureChunk feeds the analyzer with.
+  const getLastSpeechAt = useCallback(() => analyzerRef.current.lastSpeechAt, [])
+
+  return { resetMeter, measureChunk, getLastSpeechAt }
+}

src/renderer/src/hooks/use-audio-capture.ts

@@ -1,13 +1,6 @@
 import { useRef, useCallback } from 'react'
 import { openMicrophoneCaptureStream } from '@/components/dictation/microphone-devices'
-import {
-  DEFAULT_DICTATION_METER,
-  analyzeDictationAudioChunk,
-  createDictationMeterAnalyzerState,
-  dictationMeterStatesEqual,
-  toPublicDictationMeterState,
-  type DictationMeterState
-} from '../../../shared/dictation-audio-meter'
+import { useAudioCaptureMeter, type DictationMeterPublisher } from './use-audio-capture-meter'
 import { useAudioCaptureStopTail, type StopCaptureOptions } from './use-audio-capture-stop-tail'
 
 type BufferedAudioChunk = {
@@ -35,9 +28,6 @@ export type StartAudioCaptureResult = {
 
 const MAX_BUFFERED_AUDIO_SECONDS = 30
 const MAX_BUFFERED_AUDIO_BYTES = 8 * 1024 * 1024
-const METER_PUBLISH_INTERVAL_MS = 1000 / 15
-
-type DictationMeterPublisher = (meter: DictationMeterState) => void
 
 export function useAudioCapture(publishMeter?: DictationMeterPublisher) {
   const streamRef = useRef<MediaStream | null>(null)
@@ -54,10 +44,7 @@ export function useAudioCapture(publishMeter?: DictationMeterPublisher) {
   const capturedChunkCountRef = useRef(0)
   const sessionIdRef = useRef('desktop')
   const trackLostCleanupRef = useRef<(() => void) | null>(null)
-  const meterAnalyzerRef = useRef<ReturnType<typeof createDictationMeterAnalyzerState>>(undefined!)
-  meterAnalyzerRef.current ??= createDictationMeterAnalyzerState()
-  const publishedMeterRef = useRef(DEFAULT_DICTATION_METER)
-  const lastMeterPublishedAtRef = useRef(Number.NEGATIVE_INFINITY)
+  const { resetMeter, measureChunk, getLastSpeechAt } = useAudioCaptureMeter(publishMeter)
 
   const cleanupCaptureResources = useCallback(() => {
     trackLostCleanupRef.current?.()
@@ -84,13 +71,6 @@ export function useAudioCapture(publishMeter?: DictationMeterPublisher) {
     bufferedAudioSecondsRef.current = 0
   }, [])
 
-  const resetMeter = useCallback(() => {
-    meterAnalyzerRef.current = createDictationMeterAnalyzerState()
-    publishedMeterRef.current = DEFAULT_DICTATION_METER
-    lastMeterPublishedAtRef.current = Number.NEGATIVE_INFINITY
-    publishMeter?.(DEFAULT_DICTATION_METER)
-  }, [publishMeter])
-
   const removeOldestBufferedAudioChunk = useCallback(() => {
     const chunk = bufferedAudioRef.current.shift()
     if (!chunk) {
@@ -190,23 +170,7 @@ export function useAudioCapture(publishMeter?: DictationMeterPublisher) {
             return
           }
           const samples = new Float32Array(e.inputBuffer.getChannelData(0))
-          const now = performance.now()
-          meterAnalyzerRef.current = analyzeDictationAudioChunk(
-            samples,
-            now,
-            meterAnalyzerRef.current
-          )
-          if (
-            !document.hidden &&
-            now - lastMeterPublishedAtRef.current >= METER_PUBLISH_INTERVAL_MS
-          ) {
-            lastMeterPublishedAtRef.current = now
-            const meter = toPublicDictationMeterState(meterAnalyzerRef.current)
-            if (!dictationMeterStatesEqual(publishedMeterRef.current, meter)) {
-              publishedMeterRef.current = meter
-              publishMeter?.(meter)
-            }
-          }
+          measureChunk(samples)
           capturedChunkCountRef.current += 1
           if (bufferAudioRef.current) {
             appendBufferedAudioChunk({
@@ -277,7 +241,7 @@ export function useAudioCapture(publishMeter?: DictationMeterPublisher) {
     [
       appendBufferedAudioChunk,
       cleanupCaptureResources,
-      publishMeter,
+      measureChunk,
       resetBufferedAudio,
       resetMeter
     ]
@@ -336,6 +300,7 @@ export function useAudioCapture(publishMeter?: DictationMeterPublisher) {
     flushBufferedAudio,
     discardBufferedAudio,
     getCapturedChunkCount,
+    getLastSpeechAt,
     isCapturingRef
   }
 }

src/renderer/src/i18n/locales/de.json

@@ -16970,7 +16970,8 @@
           "5d2c3e7ae3": "Keine Sprache erkannt.",
           "micFallback": "Ausgewähltes Mikrofon nicht verfügbar. Systemstandard wird verwendet.",
           "micDisconnected": "Mikrofon getrennt. Diktat gestoppt.",
-          "cloudLimitReached": "Diktat am 10-Minuten-Limit der Cloud gestoppt."
+          "cloudLimitReached": "Diktat am 10-Minuten-Limit der Cloud gestoppt.",
+          "stoppedAfterSilence": "Nach einer Minute Stille beendet"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "Das Diktat konnte nicht eingefügt werden und wurde in die Zwischenablage kopiert.",

src/renderer/src/i18n/locales/en.json

@@ -17355,7 +17355,8 @@
           "5d2c3e7ae3": "No speech detected.",
           "micFallback": "Selected microphone unavailable. Using system default.",
           "micDisconnected": "Microphone disconnected. Dictation stopped.",
-          "cloudLimitReached": "Dictation stopped at the 10-minute cloud limit."
+          "cloudLimitReached": "Dictation stopped at the 10-minute cloud limit.",
+          "stoppedAfterSilence": "Stopped after a minute of silence"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "Dictation couldn't be inserted, so it was copied to the clipboard.",

src/renderer/src/i18n/locales/es.json

@@ -16697,7 +16697,8 @@
           "5d2c3e7ae3": "No se detectó ninguna voz.",
           "micFallback": "Micrófono seleccionado no disponible. Se usará el predeterminado del sistema.",
           "micDisconnected": "Micrófono desconectado. Dictado detenido.",
-          "cloudLimitReached": "Dictado detenido al llegar al límite de 10 minutos en la nube."
+          "cloudLimitReached": "Dictado detenido al llegar al límite de 10 minutos en la nube.",
+          "stoppedAfterSilence": "Se detuvo tras un minuto de silencio"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "No se pudo insertar el dictado, así que se copió al portapapeles.",

src/renderer/src/i18n/locales/fr.json

@@ -16922,7 +16922,8 @@
           "5d2c3e7ae3": "Aucune parole détectée.",
           "micFallback": "Micro sélectionné indisponible. Utilisation du micro par défaut du système.",
           "micDisconnected": "Micro déconnecté. Dictée arrêtée.",
-          "cloudLimitReached": "Dictée arrêtée à la limite cloud de 10 minutes."
+          "cloudLimitReached": "Dictée arrêtée à la limite cloud de 10 minutes.",
+          "stoppedAfterSilence": "Arrêt après une minute de silence"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "La dictée n'a pas pu être insérée, elle a donc été copiée dans le presse-papiers.",

src/renderer/src/i18n/locales/ja.json

@@ -16294,7 +16294,8 @@
           "5d2c3e7ae3": "音声が検出されませんでした。",
           "micFallback": "選択したマイクが使用できません。システムのデフォルトを使用します。",
           "micDisconnected": "マイクが接続されていません。ディクテーションが停止しました。",
-          "cloudLimitReached": "クラウドの10分の上限に達したため、ディクテーションが停止しました。"
+          "cloudLimitReached": "クラウドの10分の上限に達したため、ディクテーションが停止しました。",
+          "stoppedAfterSilence": "1分間音声がなかったため停止しました。"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "音声入力を挿入できなかったため、クリップボードにコピーしました。",

src/renderer/src/i18n/locales/ko.json

@@ -16294,7 +16294,8 @@
           "5d2c3e7ae3": "음성이 감지되지 않았습니다.",
           "micFallback": "선택한 마이크를 사용할 수 없습니다. 시스템 기본값을 사용합니다.",
           "micDisconnected": "마이크 연결이 끊어졌습니다. 받아쓰기가 중지되었습니다.",
-          "cloudLimitReached": "클라우드 10분 제한에 도달해 받아쓰기가 중지되었습니다."
+          "cloudLimitReached": "클라우드 10분 제한에 도달해 받아쓰기가 중지되었습니다.",
+          "stoppedAfterSilence": "1분 동안 말소리가 없어 중지했습니다."
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "받아쓰기를 삽입할 수 없어 클립보드에 복사했습니다.",

src/renderer/src/i18n/locales/pt.json

@@ -19224,7 +19224,8 @@
           "5d2c3e7ae3": "Nenhuma fala detectada.",
           "micFallback": "O microfone selecionado não está disponível. Usando o padrão do sistema.",
           "micDisconnected": "Microfone desconectado. O ditado foi interrompido.",
-          "cloudLimitReached": "O ditado foi interrompido no limite de 10 minutos da nuvem."
+          "cloudLimitReached": "O ditado foi interrompido no limite de 10 minutos da nuvem.",
+          "stoppedAfterSilence": "Parou após um minuto de silêncio"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "Não foi possível inserir o ditado, então ele foi copiado para a área de transferência.",

src/renderer/src/i18n/locales/ru.json

@@ -17468,7 +17468,8 @@
           "5d2c3e7ae3": "Речь не обнаружена.",
           "micFallback": "Выбранный микрофон недоступен. Используется системный микрофон по умолчанию.",
           "micDisconnected": "Микрофон отключён. Диктовка остановлена.",
-          "cloudLimitReached": "Диктовка остановлена: достигнут 10-минутный облачный лимит."
+          "cloudLimitReached": "Диктовка остановлена: достигнут 10-минутный облачный лимит.",
+          "stoppedAfterSilence": "Остановлено после минуты тишины"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "Не удалось вставить диктовку, поэтому она скопирована в буфер обмена.",

src/renderer/src/i18n/locales/tr.json

@@ -16954,7 +16954,8 @@
           "5d2c3e7ae3": "Konuşma algılanmadı.",
           "micFallback": "Seçili mikrofon kullanılamıyor. Sistem varsayılanı kullanılıyor.",
           "micDisconnected": "Mikrofon bağlantısı kesildi. Dikte durduruldu.",
-          "cloudLimitReached": "Dikte, 10 dakikalık bulut sınırında durduruldu."
+          "cloudLimitReached": "Dikte, 10 dakikalık bulut sınırında durduruldu.",
+          "stoppedAfterSilence": "Bir dakika sessizlikten sonra durduruldu"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "Dikte eklenemedi, bu yüzden panoya kopyalandı.",

src/renderer/src/i18n/locales/zh-Hant.json

@@ -16743,7 +16743,8 @@
           "5d2c3e7ae3": "未偵測到語音。",
           "micFallback": "所選麥克風不可用。使用系統預設值。",
           "micDisconnected": "麥克風已中斷連線。聽寫已停止。",
-          "cloudLimitReached": "已達 10 分鐘雲端上限。聽寫已停止。"
+          "cloudLimitReached": "已達 10 分鐘雲端上限。聽寫已停止。",
+          "stoppedAfterSilence": "靜音一分鐘後已停止"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "無法插入聽寫內容，已複製到剪貼簿。",

src/renderer/src/i18n/locales/zh.json

@@ -16294,7 +16294,8 @@
           "5d2c3e7ae3": "未检测到语音。",
           "micFallback": "所选麦克风不可用。使用系统默认值。",
           "micDisconnected": "麦克风已断开。听写停止了。",
-          "cloudLimitReached": "已达 10 分钟云端上限。听写已停止。"
+          "cloudLimitReached": "已达 10 分钟云端上限。听写已停止。",
+          "stoppedAfterSilence": "静音一分钟后已停止"
         },
         "dictation-transcript-delivery": {
           "copiedAfterFailedInsert": "无法插入听写内容，已复制到剪贴板。",

src/shared/dictation-audio-meter.ts

@@ -10,6 +10,8 @@ export type DictationMeterAnalyzerState = DictationMeterState & {
   noiseFloor: number
   smoothedLevel: number
   clippingUntil: number
+  /** The `now` of the last chunk measured as speech; 0 until then. */
+  lastSpeechAt: number
 }
 
 export const DEFAULT_DICTATION_METER: DictationMeterState = {
@@ -32,7 +34,8 @@ export function createDictationMeterAnalyzerState(): DictationMeterAnalyzerState
     lastUpdatedAt: 0,
     noiseFloor: 0.008,
     smoothedLevel: 0,
-    clippingUntil: Number.NEGATIVE_INFINITY
+    clippingUntil: Number.NEGATIVE_INFINITY,
+    lastSpeechAt: 0
   }
 }
 
@@ -71,16 +74,18 @@ export function analyzeDictationAudioChunk(
   const smoothing = rawLevel > previous.smoothedLevel ? 0.58 : 0.2
   const smoothedLevel = previous.smoothedLevel + (rawLevel - previous.smoothedLevel) * smoothing
   const clippingUntil = peak >= CLIPPING_THRESHOLD ? now + CLIPPING_HOLD_MS : previous.clippingUntil
+  const isSpeaking = smoothedLevel >= 0.1 || peak >= 0.18
 
   return {
     level: smoothedLevel,
     peak,
-    isSpeaking: smoothedLevel >= 0.1 || peak >= 0.18,
+    isSpeaking,
     isClipping: now <= clippingUntil,
     lastUpdatedAt: now,
     noiseFloor,
     smoothedLevel,
-    clippingUntil
+    clippingUntil,
+    lastSpeechAt: isSpeaking ? now : previous.lastSpeechAt
   }
 }
 

src/shared/dictation-silence-stop.ts

@@ -0,0 +1,24 @@
+const DICTATION_SILENCE_STOP_MS = 60_000
+
+/**
+ * Calls onSilence once a minute passes with no speech; returns a cancel.
+ * Timing runs on now(), not on audio, so a stalled stream still counts as silence.
+ */
+export function watchDictationSilence(
+  lastSpeechAt: () => number,
+  now: () => number,
+  onSilence: () => void
+): () => void {
+  const startedAt = now()
+  let timer: ReturnType<typeof setTimeout> | undefined
+  const check = (): void => {
+    const remainingMs = Math.max(startedAt, lastSpeechAt()) + DICTATION_SILENCE_STOP_MS - now()
+    if (remainingMs > 0) {
+      timer = setTimeout(check, remainingMs)
+    } else {
+      onSilence()
+    }
+  }
+  check()
+  return () => clearTimeout(timer)
+}