> ## 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 images

> Generate high-resolution images from a prompt, or edit from references. Lite, Standard and Pro quality. Identity preservation across up to 5 references in Pro.

Image mode generates vertical 9:16 images from a prompt, or edits
from reference images.

## Variants

| Variant            | What you bring             | For                            |
| ------------------ | -------------------------- | ------------------------------ |
| **Text to image**  | A prompt                   | Create an image from scratch.  |
| **Image to image** | 1 to 3 references + prompt | Reinterpret or combine images. |

## Quality

* **Lite**: fast and cheap for iterating.
* **Standard**: balanced for most cases.
* **Pro**: max fidelity and best identity preservation when there are
  references.

## Cost estimate

| Quality      | Cost per image |
| ------------ | -------------- |
| **Lite**     | 2 cr           |
| **Standard** | 4 cr           |
| **Pro**      | 6 cr           |

Example: 4 Pro candidates = 24 cr. Fits \~5 series in a Starter month.

<Note>
  Each input image is capped at 5 MB. If you upload something larger,
  Zevor warns you so you can shrink it first.
</Note>

## Prompt examples

* **Text to image**: "Editorial medium shot, brunette model with
  freckles leaning on a car on a Lisbon street at sunset, warm natural
  light, looking at camera."
* **Image to image**: upload a photo + prompt "same character on a
  Madrid rooftop, sunset, neutral streetwear".

## Tips

* References win: to keep a face or a style, attach a clear image of
  it.
* Describe framing and lighting, not just the content.
* To keep the same face across series, instead of repeating the
  prompt use a saved avatar (see [AI Avatars](/en/modes/characters)).