Automatic speaker identification

Accurate speaker labels make it so much easier to review the transcript and instantly understand who said what. Our AI-powered algorithm can identify speakers automatically. Just give it an example of each speaker’s voice, and it will label the entire transcript. 

Please note:

• Speaker identification during the meeting is available only in the Colibri app for Zoom. Keep the app open during your Zoom meeting to see who is speaking.
• If you record audio from Google Meet, Webex, Whereby, or any other web-conferencing or streaming platform, automatic speaker identification will be performed after the recording has stopped.
• You can teach Colibri to recognize who is speaking by tagging at least one phrase for each speaker in the meeting.
• Which phrases you tag is extremely important. Colibri will use them as voiceprints to recognize the person in all your meetings. Please try to select phrases without background noise and speech from other participants. 

Here’s how to identify multiple speakers in your transcript:

1. Before you start recording your meeting, enter participants’ names under Meeting info.
   
   
   
2. After you've recorded your meeting, go to Colibri and select the transcript from your meeting library. In the transcript, you’ll see results of unsupervised speaker identification. We call it unsupervised because you haven’t added any speaker labels yet. Consider it a draft.
3. Check the number of speakers displayed in the banner. If the number is incorrect, click Incorrect count, enter the correct number, and click Auto-tag speakers. After processing is over, you’ll see changes in the transcript.
   
   
   
4. Tag at least one phrase for each speaker. Colibri will use it as a voiceprint to recognize this person across your conversations. The phrase should be longer than 2 seconds. Please try to select phrases without background noise and speech from other participants. 
   
5. When you’ve tagged at least one phrase for each speaker, click Auto-tag speakers. Colibri will automatically tag the rest of the transcript in seconds.
   
6. You can close the banner. To open it again, open the menu at the top right corner and choose Rematch speakers.

Here’s how to change the speaker label for a particular phrase:

Automatic speaker identification algorithm is rather accurate but not perfect. Occasionally, it may assign a wrong speaker to a particular phrase. To correct it, click on the speaker’s name and select another name from the dropdown, or add a new one.

Here’s how to edit participants’ names:

1. Click on the participants’ names in the meeting header.
2. Click on a name that you want to edit.
3. Edit the name in the pop-up and click Save.

Here’s how to remove a phrase from the voiceprint database:

When you tag a phrase with a particular speaker, it is saved in the voiceprint database and used to identify this speaker across all your conversations. So, if you made a mistake labeling a phrase, it is important to undo it. Here is how:

1. Find the phrase that you want to remove from the voiceprint database. All voiceprints are marked with a verified badge.
2. Click on the speaker’s name.
3. Click Untag.

Here’s how to delete a participant from the conversation:

1. Click on the participants’ names in the meeting header.
2. Click on the close icon next to the name that you want to remove.

Have more questions? Contact us at support@colibri.ai.