Subtitles

Video has words in it too. Subtitles turns the speech in your video into a subtitle file, translated, with the timing preserved. It runs in your browser, on your device.

Subtitles

Drop a video or an existing subtitle file (.srt or .vtt). Speech is transcribed on your device by the speech model, which is downloaded once the first time you use it from a public model host and kept locally; the page names the host and the size before it starts, and asks. Each cue is translated with its neighbours in view, so a sentence split across two cues is translated as one sentence and put back in two, and the timestamps are never touched. Export as .srt or .vtt, in the target language or as a bilingual file.

A long video takes a while — transcription is the slow part — and the status line counts cues and shows a measured estimate. Stop keeps what is finished.

Limits, honestly

• Transcription quality follows audio quality. Music under speech, crosstalk and heavy accents produce more errors; review the cues before you publish them.
• A translation that is much longer than its source can be hard to read in the cue’s time. The page shows which cues are tight so you can shorten them.
• Speaker identification is not attempted: one voice per language.
• Subtitles are in the browser app and the desktop app, not in the phone apps.

Is my video uploaded?

No. Transcription and translation both run in your browser. The only download is the speech model, once, from a public model host the page names.

Which formats?

Any video your browser can play. Subtitles in and out are .srt and .vtt.