Automatic Speaker Detection

What happened to "Speaker Detective"?

As of our Release Version 90 update, we are slowly rolling out a simplified way to add speaker labels to a transcribed file. Descript will now automatically run speaker detection on transcribed files transcribed in Descript. With this update, you don't need to run Speaker Detective manually. Simply import your file and rename the speaker labels in your script.


Don't have the update yet? Expand this section to learn more
Speaker Detective is a straightforward, AI-powered super sleuth that automatically applies speaker labels for you. Just select the number of speakers and let the AI work while you focus on more important tasks. Once it's done, just fill in the labels for each identified speaker and you're set. Descript applies the speaker labels automatically across your composition.

Speaker Detective is only supported on files between 30 minutes and 3 hours long.

Using Speaker Detection

  1. Initiate an automatic transcription for a file or files; a transcription confirmation window will appear.
  2. Select the number of speakers in the file, or click Unsure, to initiate Speaker Detection. Speaker Detection will run as your file is automatically transcribed. Typically, a 60-minute file takes 2–3 minutes, but a notification will alert you when it’s done.
  3. When the notification appears, click Identify speakers.
  4. When prompted, click the Get started button and then listen through the audio prompts to identify who’s speaking. Add the speaker's name, or select an existing name from the list.
  5. When you’ve listened to all of the audio samples, click Close and the Speaker labels will appear in your transcript.

Your speakers are now labeled. Once you add speaker labels, Descript will remember them and apply them intelligently even as you cut, copy, and paste among compositions.

How to initiate speaker detection after transcribing a file

If you didn't identify speakers when initially transcribing your file, don’t worry: you can still do it afterwards from the Project panel.

  1. Open the Project panel and make sure you're looking at the Files section.
  2. Click the Snag_1099254f.png ellipsis button next to the file.
  3. Select Detect speakers.
  4. Select the number of speakers in your file, or click Unsure. Speaker Detection will then start to run.
File Size Limit for Automatic Speaker Detection

Please note that our new automatic 'Detect Speakers' feature supports files up to 5 hours long. For files exceeding this limit, the option will be disabled and can be identified by a tooltip explaining the time restriction upon right-clicking the file.

Was this page helpful?
0 out of 0 found this helpful