Supports files of any length. Each file is one session: re-picking it resumes where you left off per silence-sensitivity, and extending the range reuses what's already decoded and analysed instead of redoing it.
Remembered files
📄
Runs entirely in your browser — audio is sent directly to huggingface.co, never through formanter.py.
The token is stored only in this browser's local storage. Per-sentence transcription only —
"Transcribe whole slice" needs the local backend for segment timestamps. If the model shows as
"loading", the free tier is warming it up; try again in ~20s. If it 404s, pick a different model
id from huggingface.co's ASR models.
Select the time range to load & analyse
0%
Silence sensitivity (pause length that splits a sentence):300 ms
0 sentences detected▶ reference · 🐌 slow · ● record · ⇄ compare · 📝 transcribeneeds Local backend — switch it above ↑
Sessions (1 file = 1 session; one analysis per silence-sensitivity value)