Transcription¶
Transcription turns selected audio into timed captions on a new caption track.
Create captions¶
Select one or more audio clips, or select an audio track.
Open the timeline context menu and choose Transcribe.
Choose one of the speech-to-text models offered by the compute server.
Select Transcribe and wait for the operation to finish.
The current server can offer Parakeet, Qwen3 ASR, Whisper, and Distil-Whisper. If none appear, check the selected server in .
Follow edit points¶
Keep Follow cuts enabled when caption boundaries should follow nearby audio or video edits. Snap source chooses which cuts to follow, while Snap tolerance controls how close a generated boundary must be before it moves to a cut.
The defaults are suitable for most projects. Disable Follow cuts when you want the transcription model’s continuous timing without edit-based chunking.
If no speech is detected, Shrimply leaves the project unchanged.