VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
Python aiaudiobookcudadubbingelevenlabs-alternativehuggingfacelocal-firstmlxomnivoice-studiospeech-to-texttext-to-speechtranscriptiontranslatettsvoice-aivoice-cloningvoice-generationvoicestudioworkflow
fix(dictation): ignore empty utterance finals so pauses add no spaces (#3154)
velixiofc916b65476b
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex d31eadf29..fd24a6355 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -19,6 +19,7 @@ metadata and the backend fallback mirror it.### Fixed+- Dictation no longer inserts runs of spaces when you pause mid-sentence (#3154)- Older synthesized takes play again with accurate play/pause controls after a newer generation (#3102)- Generation-complete notifications appear above the workspace instead of covering synthesis controls (#3102)- Generated speech no longer starts and ends with silence, and `[pause]` markers give exactly the pause you asked for (Raw preset and explicit durations untouched) (#3080)diff --git a/electron/src/renderer/src/features/transcriptions/live-dictation.test.ts b/electron/src/renderer/src/features/transcriptions/live-dictation.test.tsindex 5fea8d9e8..5ad0e557c 100644--- a/electron/src/renderer/src/features/transcriptions/live-dictation.test.ts+++ b/electron/src/renderer/src/features/transcriptions/live-dictation.test.ts@@ -99,6 +99,19 @@ it('streams PCM and preserves repeated utterances without duplicating the EOF suexpect(stopTrack).toHaveBeenCalled();expect(stopGraph).toHaveBeenCalled();});+it('ignores empty utterance finals so pauses add no extra spaces (#3154)', async () => {+ const result = vi.fn();+ await live.start(result);+ const socket = Socket.instances[0];+ socket.frame({ type: 'final', final_kind: 'utterance', text: 'First sentence.' });+ socket.frame({ type: 'final', final_kind: 'utterance', text: '' });+ socket.frame({ type: 'final', final_kind: 'utterance', text: ' ' });+ socket.frame({ type: 'partial', text: ' ' });+ expect(live.getSnapshot().text).toBe('First sentence.');+ socket.frame({ type: 'final', final_kind: 'utterance', text: ' Second sentence. ' });+ expect(live.getSnapshot().text).toBe('First sentence. Second sentence.');+ expect(result).toHaveBeenCalledTimes(2);+});it('releases late microphone permission after cancellation', async () => {let allow!: (value: { getTracks: () => { stop: typeof stopTrack }[] }) => void;microphone.mockImplementation(diff --git a/electron/src/renderer/src/features/transcriptions/live-dictation.ts b/electron/src/renderer/src/features/transcriptions/live-dictation.tsindex 94da33dcf..c5dcc3114 100644--- a/electron/src/renderer/src/features/transcriptions/live-dictation.ts+++ b/electron/src/renderer/src/features/transcriptions/live-dictation.ts@@ -236,7 +236,7 @@ export class LiveDictation {}if (message.type === 'partial') {this.publish({- text: [...this.committed, message.text || ''].filter(Boolean).join(' '),+ text: [...this.committed, message.text || ''].filter((part) => part.trim()).join(' '),});return;}@@ -245,7 +245,10 @@ export class LiveDictation {const kind = message.final_kind === undefined ? 'summary' : classifySherpaFinal(message);const text = message.refined_text || message.text || '';if (kind === 'utterance') {- this.committed.push(text);+ // Silence arrives as empty finals; committing them would add a separator per pause (#3154).+ const clean = text.trim();+ if (!clean) return;+ this.committed.push(clean);this.publish({ text: this.committed.join(' ') });try {this.onResult(message);