Transcription

Selected audio becomes timed captions on a new caption track.

Create captions

1. Select one or more audio clips, or select an audio track.

2. Open the timeline context menu and choose Transcribe.

3. Choose a speech-to-text model. The server can offer Parakeet, Qwen3 ASR, Whisper, and Distil-Whisper.

4. Select Transcribe and wait for it to finish.

If no models appear, check the server in Preferences ‣ External.

Follow edit points

Follow cuts snaps a caption boundary to a nearby edit. Snap source chooses which edits count. Snap tolerance is how close the boundary has to be.

Turn Follow cuts off to keep the model’s own timing.

If no speech is found, the project stays unchanged.