How it works
- Drop an MP4, MOV or WebM file on the box.
- Leave language on auto, or pick one. Tick the translate box to get English captions from any language.
- Press Make captions, fix any wrong words in the list, and export.
The model is Whisper, the open speech recognition model from OpenAI, in its base size. It runs inside this tab using your graphics chip when the browser allows it and plain CPU code when it does not. The first run downloads about 80 MB of model files and your browser keeps them, so the second video starts at once.
Speed depends on your machine: a laptop with WebGPU listens faster than the clip plays, and a CPU only machine can take longer than the clip. A 3 minute video is the free limit. Paid plans go to 30 minutes.
Whisper gets names and jargon wrong now and then. That is why the caption list is editable before you export: click a line, retype the word, and the SRT and the burned-in video both use your fix.
Questions
Is the video uploaded?
No. The file is read in your tab, the speech model runs in your tab, and the captions are built in your tab. Only the payment check for paid plans talks to our server.
Which languages work?
Whisper handles about 99 languages. Auto detect picks one from the first seconds of audio. Choose the language yourself if the clip opens with music.