Polish or redo the transcript with AI

You can polish the transcript with AI, redo it, and re-run speaker diarization. This page explains the operations that improve quality.

Last updated: October 1, 2026

Polish the transcript with AI

AI polish cleans up misrecognitions and spelling variants in one pass. The pre-polish result is kept, and you can always revert with Reset polish.

1. Open the meeting and show the Transcript tab.
2. Press AI polish to run the polish.
3. Review the result and use Reset polish to revert if needed.

Redo the transcript

If you want a different model or the result is not good, use Redo transcript to regenerate.

You can switch between local and cloud transcription with Transcribe locally or Transcribe in the cloud.

Re-run speaker diarization

Diarization splits utterances by speaker and adds labels. You can run it after transcription, or change settings and re-run it.

Local diarization identifies speakers with an on-device model and never sends audio off the device. Cloud diarization uses more allowance than regular transcription.

You can assign names to the separated speakers afterward.

Trim the recording

If there is unnecessary audio at the start or end, use Trim recording to cut it.

Review the start and end of the transcript, exclude the unnecessary range, then confirm.