Skip to main content

Speaker Separation (Diarization)

Automatically identify and separate different speakers in your audio.

What is Speaker Diarization?​

Diarization automatically:

  • Detects when different speakers are talking
  • Separates speech into distinct speaker turns
  • Groups speech by the same speaker
  • Labels speakers with identifiers

When to Use It​

Enable speaker separation for:

  • Interviews
  • Meetings and conference calls
  • Podcasts with multiple hosts or guests
  • Panel discussions
  • Debates
  • Customer service calls
  • Any content with 2 or more speakers

Don't enable for:

  • Solo presentations or lectures
  • Single-person podcasts
  • Voice notes
  • Any single-speaker content

How to Enable​

During upload, find the "Separate speakers" toggle and turn it ON. The setting applies to that upload only.

What You Get​

With diarization: separate utterances per speaker turn, speaker labels, ability to rename. Without: continuous text blocks, no labels, manual splitting required.

Working with Speakers​

In the editor, click speaker labels to rename, split combined utterances, or merge duplicate speakers.

Tips​

For best results:

  • Clear turn-taking between speakers
  • Minimize people talking over each other
  • Reduce background noise
  • Use good quality audio

After transcription:

  • Review speaker labels
  • Rename generic labels with actual names
  • Correct any misattributed speech
  • Then edit the transcript text

Next Steps​