Free tool
Transcribe video & audio locally
Drop a file in. Nothing leaves your machine.
- No upload
- No account
- Unlimited length
- SRT & VTT
How it works
- 1
Whisper downloads
OpenAI's speech model, fetched into your browser once and cached. ~45 MB.
- 2
Your GPU runs it
WebGPU does the inference on your own graphics card. No server is involved.
- 3
Text comes back
Copy it, or export SRT and VTT with timestamps for a video editor.
After the first load you can turn off your wifi and it still works. Plenty of products charge a subscription to wrap something that now runs free on your own hardware.
Questions
Is my file uploaded anywhere?
No. The audio never leaves your browser. There is no server to upload it to — this site is static files.
What formats can I drop in?
Anything your browser can play: MP4, MOV, WebM, MP3, M4A, WAV, FLAC, OGG. MKV and AVI usually fail, because browsers ship no decoder for them — remux to MP4 first and it will go through.
Why is it slow in Firefox?
Firefox's WebGPU support is still patchy, so the page falls back to WebAssembly on the CPU. It works, but expect several times the wait. Chrome, Edge and Safari 26+ use the GPU.
How long a file can it handle?
There is no limit built in. Memory is the ceiling — an hour is fine on a modern laptop. Very long files can drift slightly on timestamps toward the end. That is a known Whisper behaviour, not something this page introduces.
Can it tell speakers apart?
Not yet. Speaker diarization needs a second model, and the in-browser options for it are not good enough to ship.
A few hundred lines. No backend.
That is what the course teaches — building fast, and getting it in front of people.