← All models
Commercially safe, multi-reference image editing model. Follows natural language instructions alone or with up to 4 reference images, purpose-built for complex object and character combinations, virtual try-on, background replacement, style transfer, and more.
images
Details
- Accepts
- text
Pricing
- Price
- 5 cr / image
Prices in credits (1 credit = $0.01).
Data schema
Input
| Field | Type | Description |
|---|---|---|
| seed | integer | Random seed for reproducibility. |
| mask_url | — | Mask (file or URL) marking the region to regenerate: white where the model should edit, black elsewhere. Single-reference requests only, and it must be the same size as that image. A masked edit comes back at the reference's own resolution. |
| sync_mode | boolean | If true, returns the image directly in the response (increases latency). |
| image_urls | — | 1-4 reference images (files or URLs). Order is significant: the instruction is resolved against the images in the order they are sent. |
| instruction | — | Instruction for image editing. |
| aspect_ratio | — | Output aspect ratio. Left unset, the output keeps the ratio of the first reference image. A chosen ratio applies only with two or more reference images; with a single reference the output keeps that image's ratio either way. |
| structured_instruction | — | A pre-built structured prompt, used verbatim instead of having VGL build one when no instruction is sent. Accepts what a previous edit returned. |
Output
| Field | Type | Description |
|---|---|---|
| image | — | Generated image. |
| images | array | Generated images. |
| structured_instruction | object | Current instruction. |