> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zevor.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Avatars

> Generate an AI avatar once and reuse it across motion clips, videos and talking photos. Same face, different poses, outfits and scenes.

An AI Avatar is a consistent character: same face, distinctive marks
and body across generations. You create it once, save it to your
library, and reuse it in any mode.

## Create an avatar

<Steps>
  <Step title="Describe the character">
    Use the attributes (chips) to build the description, or write your
    own free-text prompt.
  </Step>

  <Step title="Generate and pick">
    The AI generates 4 candidates. Keep the one you like best.
  </Step>

  <Step title="Save">
    Name it and it's added to your avatar library to reuse. You can
    also **Download** the face directly.
  </Step>
</Steps>

## Cost estimate

| Action                        | Cost                        |
| ----------------------------- | --------------------------- |
| Create avatar (4 candidates)  | \~16 cr                     |
| Remix variations (4 images)   | \~12 cr                     |
| Animate in Motion             | 6 cr/s                      |
| Talking photo with the avatar | 4 cr per real speech second |

## "Prompt only" mode

By default, if you don't pick chips, Zevor adds a random scene,
outfit and pose for variety and realism. If you want your text to be
the only thing driving it, turn on **Prompt only**:

* Uses just your prompt plus the core photorealism (real skin, no
  CGI, 9:16 framing with no cropping).
* Adds no random scene, outfit or pose.
* Ignores the builder attributes.

<Note>
  In "Prompt only" mode you need to type text: a lone chip isn't
  enough because the builder is ignored.
</Note>

## Reuse

From the avatar page you can:

* **Remix**: same face, different pose, outfit or scene.
* **Animate**: take it into [Motion](/en/modes/motion) as the character.
* **Talking photo**: make it say whatever you want.
* **Download**: the face, or all variations as a ZIP.

## Full workflow

To take an avatar from zero to a publication series, follow the
[Build an AI influencer](/en/recipes/ai-influencer) recipe. It covers
the whole flow: base face, variations, videos and voiceovers.