Skip to main content

Video translation and dubbing

Video projects take a video (or audio) file through transcription, subtitle translation, and optionally full AI dubbing — all inside the Video Editor.

The video editor

The pipeline​

  1. Transcribe — start from the Transcribe page (see Media tools for the upload options: speaker separation, background separation, subtitle length limits). Speech-to-text runs with word-level timestamps; source language can be auto-detected, and you can hint the expected number of speakers. Re-transcribe at any time if the first pass needs another go.

  2. Translate — translate the subtitle segments with Start Translation; re-runs can optionally re-translate already translated segments. You can also import existing subtitles (SRT) instead of transcribing.

    For full review workflows, use Send to Translation Editor (in the translate dialog): the subtitles — or dubbing lines, depending on the selected mode — open as a session in the main translation editor, with complete TM/TB leverage, comments, locking, and team collaboration. Run pre-translation here first for subtitle/dubbing context-aware pre-processing, then refine in the editor. Timestamps are preserved via SRT round-tripping. When review is done, Pull Translations from Editor applies the edited text to the video.

  3. Edit subtitles — the Subtitles tab lists segments synced to the video player and timeline. Edit text and timings, and split a segment at the playhead or join it with the next. Undo/redo and keyboard navigation (arrow keys, space to play/pause) make long sessions workable.

  4. Dub — the Dubbing tab generates AI voice-over (ElevenLabs) from the translated segments.

  5. Export — the finished video with dubbed audio, audio only, or subtitle files.

Dubbing​

Dubbing works on blocks — groupings of subtitle segments that are voiced as one continuous utterance:

  • Dub all segments generates audio for everything; you can also dub or re-dub individual segments and blocks.
  • Regenerate to fit — when translated speech runs longer than the original slot, regenerate with a target pacing (e.g. much shorter (0.7x) or same length (1.0x)) so the dub stays in sync.
  • Speakers and voices — assign a voice per speaker, reassign speakers where diarisation got it wrong, and clone voices for a closer match to the original.
  • Filler-word muting — analyse the original audio for filler words and mute them.
  • The original audio track is always preserved alongside generated tracks.

Blocks can be split, merged, or cleared and rebuilt without losing the underlying subtitle segments.

Timeline editing​

The timeline shows subtitle and dubbing blocks against the audio. Drag to move, resize to retime — blocks snap magnetically to neighbouring boundaries and the playhead for clean alignment.

:::tip Who does what PMs typically run transcription, translation, and dubbing; linguists work in the Subtitles tab refining text and timings like any other editor session. :::