How to Use Speaker Diarization

Speaker diarization automatically figures out how many people are speaking in a recording and labels each transcript segment with the right speaker — instead of one undifferentiated wall of text.

Turning it on

1. In Settings → General, enable the "Speaker diarization" toggle to make it the default for new transcriptions.
2. Or, on the Transcribe screen, use the "Detect speakers" option before starting a transcription to enable it for that file only.
3. Once transcription finishes, each segment shows a speaker label (Speaker 1, Speaker 2, etc.).

Renaming speakers

Generic labels like "Speaker 1" aren't always useful — click a speaker label in the finished transcript to rename it to the person's actual name. The rename applies to every segment from that speaker throughout the transcript.

Where this matters most

• Interviews — separating interviewer from subject cleanly
• Meetings — knowing who said what for accurate minutes
• Podcasts with co-hosts or guests
• Legal or research recordings where speaker attribution matters

This is a Pro feature

Diarization is included on paid plans. On the free plan, transcripts are generated without speaker separation.