Descript lets you create AI avatars that animate in sync with your speaker's voice. You can upload a photo of a human face or generate an avatar using a text prompt—no camera required. It's a flexible way to personalize your content while working within Descript's text-based editing workflow.

This article covers:

• How to create a custom avatar with an image
• How to create an AI-generated avatar from a text prompt
• How to update or change an avatar in your composition
• Best practices for using custom avatars

Create a custom avatar with an image

1. Assign a speaker to your script.
2. Click the speaker label and choose Assign avatar.
   
3. Click Upload photo and select a supported image file (.jpeg, .png, or .webp).
   
4. Preview your image and click Assign avatar to apply it.
   
5. Once your script is finalized, click Generate avatar to animate your image and sync it with the speaker.
   

If your photo doesn't meet formatting or safety requirements, Descript will prompt you to upload a different one.

Create a custom avatar from a text prompt

1. Assign a speaker to your script.
2. Click the speaker label and choose Assign avatar to [speaker name].
3. In the Text prompt tab, enter a description to generate an avatar, or click the Inspire button to use a suggested prompt. For best results, describe a realistic, human-like figure.
4. Descript will generate three avatar options. You can:
   • Select one of the images to apply, or
   • Continue iterating on your prompt. Previous generations remain visible as long as the avatar modal stays open.
   Note: Closing the modal will clear your avatar history.
5. Click Assign avatar to apply your selection.
6. Once your script is finalized, click Generate avatar to animate the image and sync it with the speaker.

Generation workflow and timing

• Avatar generation uses avatar minutes, not AI voice minutes.
• Minutes are calculated based on total spoken audio duration.
• You’ll see a modal showing your remaining avatar minutes and estimated usage before confirming.
• The current max length of an avatar generation is 12 minutes.
  Avatar generation continues in the background if you close the project. You’ll receive an email when it’s ready.

Managing and updating custom avatars

To update a custom avatar, click the speaker label and choose Update speaker’s avatar. This updates the avatar across your project while preserving its size, position, and visibility in the scene editor.

If the avatar doesn’t appear in the scene:

• Open the Scene panel and click the "show layer" icon
• Drag the avatar layer above other visuals if needed

Or use Replace media in the scene editor to swap it back into view.

Avatar layers aren’t visible in the timeline but can be cropped, styled, or repositioned in the scene editor. You can also apply effects like greenscreen.

Best practices for using custom avatars

Whether you’re uploading a photo or generating an avatar from a text prompt, use a clear, human-like face with even lighting and a relaxed, natural expression. Avoid using images or prompts that include animals, objects, or abstract characters—they won’t animate reliably.

Head position and framing

Glasses, mouth, and eyes

Background, foreground, and lighting

File upload requirements