On current plans, this feature uses AI credits.
1
Get your content into Descript
Record new content, generate it with text-to-speech, or import an already captured file.
2
Add or select a Speaker label
- No Speaker label yet: click Add speaker. It appears at the top of your script, and at the start of each section attributed to a specific Speaker.
-
Already labeled: click the existing Speaker label.

3
Open the Speaker card
- Use an existing Speaker
- Create a new Speaker
If your script already has a Speaker label, click the pencil (Edit speaker) beside the Speaker’s name to open the Speaker card.


4
Add an avatar
In the Speaker card, click Create or choose avatar.Select a stock avatar from the gallery, or create a custom avatar.Click Save to apply the avatar to the Speaker.

5
Animate the avatar
After you’ve finished editing your script, click Generate avatar.

6
Set the attitude and model
In the Generate avatar dialog, you can fine-tune how your avatar is animated before you generate it.
- Attitude: click the attitude button next to the Speaker’s name, then pick a preset (Friendly, Professional, Calm, or Excited) or describe your own. Your speech audio still drives most of the avatar’s expression.
-
Model: choose a different avatar model from the model picker.

7
Generate your avatar
Click Generate. Descript animates your avatar to sync with your speech audio.
Create a custom avatar
Create a custom avatar from a text prompt, an attached image, or both, then click Create.From a text prompt
Describe a realistic, adult, non-celebrity, human-like figure, or click Inspire for a suggested prompt. You can further influence the output by choosing a look from the style library. Descript generates three images. From there, you can:- Select the one you want to use.
- Keep iterating on your prompt.
- Restyle an image while keeping the same base: hover the preview and click Add image to prompt, then describe the change. Descript generates from both your prompt and the image.
From an image
Upload a.jpeg, .png, or .webp file up to 10 MB. Widescreen (16:9) images look best. Like AI-generated images, uploaded photos pass safety checks and may be rejected if they contain celebrities, minors, or explicit content.
For reliable animation, use a clear, human-like face with even lighting and a relaxed, natural expression. Avoid animals, objects, or abstract characters.
- Framing: a close-up headshot, centered and squared to the camera, shoulders slightly visible, with nothing blocking the face.
- Head and face: upright and forward-facing, not angled; eyes open and clearly visible; mouth unobstructed for accurate lip sync; glasses without heavy reflections.
- Background: simple, uncluttered, and well-contrasted, with no other people or animals. Avoid transparency.
- Lighting: soft and even, with no harsh shadows.
Manage avatars in your project
Once assigned, an avatar “follows” its Speaker everywhere they appear in your composition. Each Speaker can be connected to one avatar at a time. After you generate it, you can:- Change or remove it: click the pencil (Edit speaker) beside the Speaker’s name to open the Speaker card.
- Resize, reposition, or crop it in the Scene panel.
- Combine it with effects like Green Screen.

