Skip to main content
Speaker labels help you follow who says what in a meeting recording. Use matching colors to connect people with their transcript turns. If you own the recording, you can correct an identity. When speaker identification is available, you can also enroll your own voice through the API so future recordings can recognize you.

How to match speaker colors and name or correct a speaker

  1. Review the speakers
  2. Match the colors
  3. Assign an unnamed speaker
  4. Search people
  5. Pick the person
  6. Open a mistaken speaker
  7. Rule the person out

Athena 101 lesson: Turn a meeting into decisions (broader walkthrough)

​
Follow speakers in a recording

  1. Open the meeting recording and review the people listed in Overview. A speaker with no name yet shows a link to assign them in Recording.
  2. Match each speaker’s color dot with the same color beside their transcript turns.
  3. Open Recording to read the transcript and review speaker names.
Colors help you follow a speaker within that recording. They do not replace checking the name and listening to the audio.

​
Correct a speaker identity

  1. As the recording owner or a workspace owner, open Recording.
  2. Select Assign beside an unnamed speaker in the People card, or select a speaker label in the transcript. The label is a button titled Name this speaker.
  3. Type in the Who is speaking? field to find the correct person. Choose This is me for your own voice, or select a matching person.
  4. If you need a new person, select New person… and enter their name.
  5. To rule out a mistaken match, select the option beginning with Not followed by that person’s name.
Assigning a person updates every transcript line for that voice. It can also contribute confirmed voice evidence for future identification. Ruling a person out removes the mistaken assignment and prevents later identification runs from naming that person for that voice.

​
Enroll and review your own voice through the API

  1. Query voiceProfileEnrollmentInfo to check availability and accepted sample limits.
  2. Record only your own voice. Include at least 30 seconds of detected speech. Keep the recording within five minutes and 50 MB.
  3. Submit a multipart request to /api/speaker-identity/enroll-sample with file and, optionally, device_label.
  4. Query myVoiceProfile for your profile summary. Use mySpeakerRecordings and myEnrollmentClips to review its contributing recordings and enrollment clips.
  5. Use withdrawVoiceSession(sessionId: ...) to remove one of your own contributions.

​
Limits

Enrollment requires active workspace membership and speaker-identification support. You can enroll only your own voice. Withdrawing a contribution rebuilds your profile and removes its enrollment clip when applicable. It does not delete the source meeting recording. Speaker identity does not grant access to assets. Your existing recording permissions still apply. The profile steps above use API fields, rather than a customer settings panel.