Use Regenerate to change what was said

Use Regenerate to replace or rewrite parts of your script using an authorized custom AI speaker. The regenerated audio and video stays in sync with your script edits. Use it when you misspoke, want to update a phrase, or need to fix a line without going back to the mic.

If you want to smooth pacing or heal jump cuts without changing what was said, see Use Regenerate to smooth or improve a section.

Regenerate is currently available for English audio and video only.

Before getting started

To replace what was said, the original speaker must have a custom AI speaker created and authorized for use in Descript. If a custom AI speaker isn't authorized and assigned, Regenerate can't be used to make script changes or alter the original audio. Learn how to create a new custom AI speaker.

Change what was said with Regenerate

  1. Make sure your script has a speaker assigned.
  2. Highlight the section of script you want to change.
  3. Click Regenerate in the selection toolbar, or press D. A dialog opens with a text field for the new words.
    Regenerate selection in Descript
  4. Type the new phrase that will replace the original audio.
  5. (Optional) In video projects with a single script track, Regenerate defaults to Audio & video. To switch to Audio only, click the dropdown caret next to Regenerate and choose Audio only. Sequences don't support video regeneration.
    Regenerate dropdown menu in Descript
  6. Click Regenerate in the dialog. The script text pulses while Descript processes the change.
  7. Once it's ready, the Regenerate toolbar gives you four options:
    • Play: Preview the regenerated section.
    • Restore: Revert to the original recording.
    • Retry: Regenerate another variation.
    • Accept: Keep the result. The original video is clipped to the regenerated section, moved into the layer lane, and hidden.

First-time video regeneration

Descript asks for consent the first time you regenerate video, since the process modifies a speaker's visual performance.

Known limitations

  • Each Regenerate action is limited to 250 characters.
  • Regenerate won't run on a selection that crosses a scene boundary or includes ignored text.
  • Regenerate needs original audio surrounding the selection to work. If the section is bordered only by AI-generated speech or has no surrounding audio, Regenerate won't succeed.
  • In video projects, Audio & video isn't available for sequences or until you've given consent for video regeneration. The dropdown shows Audio & video grayed out and defaults to Audio only.
  • The original video (if applicable) is clipped to the regenerated section, moved into the layer lane, and hidden.
  • Video regeneration won't succeed on B-roll, multi-speaker recordings, or shots where the speaker's face isn't clearly visible. Descript will let you know if it can't complete the regeneration after running.