Readme
Reve 2.1
Reve 2.1 is a single model that generates and edits high-quality images. Describe what you want in words, and optionally pass one or more reference images to edit, combine, or take style and subjects from.
Under the hood it uses Reve’s v2 API, which folds the old create, edit, and remix workflows into one endpoint. That means you use the same model whether you’re generating from scratch or working from existing images.
What it does
- Text to image. Give it a prompt and get an image back.
- Editing. Pass one reference image and describe the change (“give her neon green hair”, “remove the people in the background”).
- Remixing and composition. Pass several reference images and describe how to combine them.
Referring to reference images
Reference images are an ordered list. You can point to a specific one from your prompt using <frame>N</frame>, where N is its position in the list (starting at 0).
For example, with two reference images:
Put the hat from <frame>0</frame> on the person in <frame>1</frame>
Inputs
| Input | Description |
|---|---|
prompt |
What to generate, or how to edit the reference images. |
reference_images |
Optional ordered list of up to 8 images to edit or remix. Leave empty for text to image. |
aspect_ratio |
Aspect ratio of the output. Use auto to let the model choose. |
Output
A single image.
Aspect ratios
Reve 2.1 supports a wide range of aspect ratios, from tall (1:4) to wide (4:1), plus auto, which picks a shape that fits your request.