First run grabs a ~75MB speech model, cached for next time. Add names below so tricky spellings come out right.
Drop audio file here or click to browse
MP3, WAV, M4A, OGG, FLAC — max 100MB
or
Audio Transcriber
Transcribe audio or mic recordings with local Whisper AI
FAQ
Is my audio sent to a server for transcription?
No. OpenAI's Whisper model runs inside your browser. Files and mic recordings stay on your device.
Can I record directly instead of uploading a file?
Yes, hit Record, speak, stop, and transcribe. The recording is kept in memory only and can be transcribed or discarded.
How do the summary and filler-word cleanup work?
Both run locally with plain JavaScript, the summary picks the most information-dense sentences and the cleanup strips fillers like 'um', 'uh', and 'you know'. No AI API, no cost.
Which languages does it support?
Whisper supports 90+ languages with automatic language detection; English gives the best accuracy on the small local models.
How accurate is it?
Very good for clear speech, podcasts, and meetings. Heavy accents, crosstalk, or noisy audio reduce accuracy, the larger model option helps.