v104: точная привязка слогов к аудио через word timestamps Whisper

This commit is contained in:
“Naeel”
2026-05-23 14:39:54 +04:00
parent 7302d6ef62
commit 6fe6f77c07
5 changed files with 236 additions and 18 deletions
+8 -4
View File
@@ -28,7 +28,7 @@ async function blobToWav(blob) {
}
async function transcribe(blob) {
if (!GROQ_API_KEY) return { text: '', error: 'no_key' };
if (!GROQ_API_KEY) return { text: '', words: [], error: 'no_key' };
const tStart = performance.now();
try {
const wavBlob = await blobToWav(blob);
@@ -36,6 +36,8 @@ async function transcribe(blob) {
fd.append('file', wavBlob, 'audio.wav');
fd.append('model', 'whisper-1');
fd.append('language', 'it');
fd.append('response_format', 'verbose_json');
fd.append('timestamp_granularities[]', 'word');
const ctrl = new AbortController();
const timer = setTimeout(() => ctrl.abort(), 30000);
const r = await fetch('https://api.proxyapi.ru/openai/v1/audio/transcriptions', {
@@ -48,11 +50,13 @@ async function transcribe(blob) {
const raw = await r.text();
console.log('[TRANSCRIBE] ← ' + r.status + ' ' + ((performance.now()-tStart)/1000).toFixed(2) + 's raw=' + raw.slice(0,200));
let data;
try { data = JSON.parse(raw); } catch(pe) { return { text: '', error: 'parse:' + raw.slice(0,80) }; }
return { text: data.text || '', error: data.error?.message || data.error || null };
try { data = JSON.parse(raw); } catch(pe) { return { text: '', words: [], error: 'parse:' + raw.slice(0,80) }; }
const words = (data.words || []).map(w => ({ word: w.word, start: w.start, end: w.end }));
console.log('[TRANSCRIBE] words:', words.length, words.map(w => w.start.toFixed(2)+'-'+w.end.toFixed(2)+' '+w.word).join(', '));
return { text: data.text || '', words, error: data.error?.message || data.error || null };
} catch(e) {
console.log('[TRANSCRIBE] ERR', e);
return { text: '', error: e.message };
return { text: '', words: [], error: e.message };
}
}