How to transcribe a video for free
- Open the file. Drop a video or audio file onto the box above, up to 10 minutes long.
- Press Transcribe to text. The first run downloads the speech model once (the size is shown first; later runs start straight away).
- Copy or download. Copy all the text, or download TXT, DOCX, SRT or VTT, with or without timestamps.
Why it can be free
Transcription usually runs on a company's servers, which is why free plans stop after a few minutes. SubForge runs the speech model inside your browser instead, on your own graphics card when it has WebGPU. There is no server bill to pass on, so there is no paywall in front of the export button.
What you get for free
| Feature | SubForge free |
|---|---|
| Sign-up | None |
| Length per file | Up to 10 minutes, as many files as you like |
| Exports | TXT, DOCX, SRT, VTT (and ASS in Subtitles view) |
| Timestamps | Optional, per paragraph; full timing in SRT and VTT |
| Watermark | None |
| Upload | None: it runs on your device |
| Speech model | Whisper large-v3 turbo where your device can run it, Whisper small or base otherwise |
How fast it is
On a computer with WebGPU, turbo transcribed our test clips in 0.4 to 0.65 times their length once the model was loaded. Without WebGPU the tool uses your CPU and a smaller model, which takes about as long as the clip. The first visit also downloads the model (564 MB for turbo); the progress bar shows the size, speed and time left.
Languages
Automatic detection covers about 99 languages. Pages with our measured accuracy: Spanish, Arabic, Urdu.
Questions
Is it really free, with no catch?
Yes. No account, no credit card, no watermark, no limit on how many files. The one limit is 10 minutes per file, so the tool stays responsive on ordinary computers.
How accurate is the free transcription?
It uses the same open Whisper models paid services use. With turbo we measured 2% word error on Spanish and 4.5% character error on Arabic read speech. Clear audio with one speaker comes out best; read the text before you publish it.
Do I need to install anything?
No. It runs in Chrome, Edge and other modern browsers. Chrome and Edge on a computer use WebGPU and are fastest.
Why does the first run take longer?
The speech model downloads once and is kept in your browser. After that, transcription starts straight away, even offline.
Can I transcribe audio files too?
Yes: MP3, WAV, M4A, AAC, OGG, Opus and FLAC work the same way.
Can I edit the transcript?
Yes. Switch to Subtitles to fix any line; the transcript follows your edits.