How it works
The browser decodes the file's audio track locally, resamples it to 16 kHz, and runs Whisper (base, quantised) through WebAssembly. The markdown is assembled in the page — there is no backend processing step to send anything to.
Turn a video into a markdown transcript.
Drop a video (or audio) file below. The audio is extracted and transcribed in this browser tab — nothing is uploaded, no account, no server-side copy.
Drop a video here
MP4, MOV, WebM, MKV, M4A, MP3, WAV · nothing is uploaded
Working…
0%
- Extract audio
- Load model
- Transcribe
First run downloads the speech model (~77 MB, cached afterwards).
Your markdown
What to expect
Transcription runs faster than real time on a laptop, slower on a phone. There is no file-size cap: very long videos are limited only by the memory of your tab, so on a low-memory device split them first.
Accuracy
A small on-device model: clear speech transcribes well, heavy accents and music less so. Names and numbers are the usual weak spots — skim before publishing.