Local speech transcription
Transcribe audio directly in your browser with Whisper and WebAssembly. Your media never leaves this device.
Help keep private transcription available to everyone
Whisper Online has no accounts, uploads, or usage fees. Your contribution helps cover development and keeps the project focused on a fast, private transcription experience.
A focused transcription workspace
From audio to an editable transcript
Transcribe, clean up individual words, and export timed captions without sending your media to a server.
Private by default
Audio stays on your device while Whisper runs locally in the browser.
Token-level editing
Correct, remove, merge, and undo individual timed words without losing their place.
Useful exports
Download subtitles, plain text, or detailed timing data when the transcript is ready.
How it works
A short path to clean captions
- 01
Choose your media
Select an audio or video file and a Whisper model suited to the speed and accuracy you need.
- 02
Transcribe locally
WebAssembly runs Whisper on your device. The selected media is decoded and processed without an upload step.
- 03
Edit and export
Refine individual tokens, merge punctuation, undo changes, and download the result in the format you need.
Flexible output
One transcript, several destinations
Create subtitles for players and editors, readable text for notes, or structured timing data for your own workflow.
Practical guides
Take the transcript into your workflow
Common questions
About local transcription
Does my media leave the browser?
No. Audio decoding and Whisper inference happen locally on your device. Model files are downloaded, but your selected media is not uploaded for transcription.
Which Whisper model should I choose?
Tiny is the quickest choice, Base balances speed and accuracy, and Small favors accuracy over download size and processing time.
Can I correct the transcript before exporting?
Yes. Every timed token is editable. Delete unwanted tokens, merge neighboring words or punctuation, and undo edits before downloading.