Labeling Speakers

Transcription Service can have your provider work out who is speaking and name them Speaker 1, Speaker 2 and so on in the transcript. It’s useful for interviews and panel shows. Leave it off for a solo show: it gains nothing, and transcribing takes longer.

Turning It On

The setting is per show, and off until you turn it on. Open the show’s Transcription settings tab and find Speaker labels:

1. Turn on Label speakers.
2. Optionally, set Number of speakers: Not sure, or 1 to 10. Only AssemblyAI uses this, and it helps stop one person being counted as two. The other providers work it out themselves.
3. Click Save speaker labels.

It applies to episodes transcribed from then on. Finished transcripts stay as they are.

Which Providers Can Do It

Under original Whisper, the toggle is replaced by Original Whisper can’t label speakers. To label them, choose gpt-4o-transcribe-diarize as your OpenAI model. and Open Settings → Transcription. Your show’s choice is kept, and returns when you change the model. A custom provider that can’t label speakers shows {Provider} can’t label speakers. With no provider chosen yet, the toggle shows.

Your provider’s model is chosen on Setting Up Transcription.

Where the Labels Appear

A labeled transcript shows its speakers in the episode’s Transcript tab and on your site, with the speaker named when it changes. Every export names them too.