Skip to main content

Voice diarization

Voice diarization separates a meeting recording by speaker and labels each line with the person who said it. In in-person meetings, where everyone shares one microphone, the agent matches each voice to room users who have recorded a voice sample. The result is a speaker-labelled transcript and AI minutes where decisions and action items carry the right names. Every user records their own voice sample.

How diarization works​

  1. Voice sample — you read a short phrase once; it becomes your voice profile.
  2. Recording — an in-person meeting is recorded in the room.
  3. Speaker separation — the agent splits the audio by voice.
  4. Speaker naming — each voice is matched to a voice profile of a room user. A voice that is not matched stays as SPEAKER 2 (or similar) until the room owner names it.

How to record your voice sample​

  1. In a room, select your own name or picture, then select Voice sample (under Speaker Diarization). You can also start it from your profile.
  2. The Voice recognition window opens. Allow microphone access if your browser asks.
  3. Select Tap to record and read the phrase on screen clearly.
  4. Wait while the sample is validated and your voice profile is verified.
  5. Select Done. If the sample is not clear enough, select Try again.

Once your voice sample is recorded, a voice badge appears beside your name in the room's attendance list.

tip

Speak at your normal volume in a quiet place. A clean sample gives the most accurate speaker identification.

How to name an unknown speaker​

  1. Open the in-person meeting's recording in the room.
  2. Select the speaker label, such as SPEAKER 2.
  3. Choose a room user, an invited user, or enter a custom name. The transcript and AI minutes update.
note

Only the room owner can assign speakers in a recording.