How it works

1. 1You paste the URLA public YouTube video, up to 20 minutes. You also pick the language to translate it into.
2. 2We look for captionsIf the video already has captions we use those — the uploader's own, or YouTube's automatic ones, in the original language.
3. 3If there are none, we listenWhen a video has no captions at all, we pull the audio and transcribe it with Whisper, which runs on the server's own machine.
4. 4We translate, only if neededIf the video is already in the language you asked for, nothing is translated and no quota is spent. Otherwise DeepL translates it.
5. 5You read along with the videoThe result comes back timed: click any line to jump there, and the line being spoken lights up as it goes.

Where it is kept

There is no database of results. The server processes the content and tries to clean up the temporary files Whisper uses when it finishes. A history of the last 5 videos is kept in this browser's local storage, with the full result inside, so you can reopen one without running anything again. The “Clear history” button removes those local copies. Translation goes through DeepL, and technical records may exist: see Privacy for how the data is handled.

Limits

• Videos up to 20 minutes.
• A limited number of transcriptions and audio minutes per session.
• Transcribing from audio takes longer than reading captions that already exist: it is a whole video going through Whisper.
• Translation uses DeepL's free plan, shared by everyone using this demo. If the monthly quota runs out, translation pauses until it renews.