Wan Image Generation & Editing
POST /api/v1/services/aigc/multimodal-generation/generation — wan2.7-image text-to-image / image sets / editing
The Wan 2.7 image model covers text-to-image, image-set generation, image editing, and multi-image reference generation. Synchronous, on the same entrance and message shape as Qwen image, with its own knobs.
Models
| Model | Notes |
|---|---|
wan2.7-image | Wan 2.7 image (faster generation) |
Create
/api/v1/services/aigc/multimodal-generation/generationcurl https://api.modelsite.ai/api/v1/services/aigc/multimodal-generation/generation \
-H "Authorization: Bearer $MODELSITE_API_KEY" -H "Content-Type: application/json" \
-d '{
"model": "wan2.7-image",
"input": {
"messages": [
{ "role": "user",
"content": [{ "text": "Ink-wash style Jiangnan water town at dawn, mist on the canal" }] }
]
},
"parameters": { "size": "2K" }
}'Request body
| Field | Type | Notes |
|---|---|---|
model | string (required) | wan2.7-image |
input.messages | array (required) | One {"role": "user", "content": [...]} (single turn) |
input.messages[].content | array (required) | {"text": "..."} (≤5000 characters, required) plus 0-9 {"image": "..."} entries (URL or Base64; sides [240, 8000]px, ≤20MB; multi-image order defines reference order) |
parameters | object (optional) | Below |
parameters
| Parameter | Type | Notes |
|---|---|---|
size | string | Two mutually exclusive styles: tiers (1K / 2K (default); pure text-to-image also 4K) or pixels W*H (asterisk; area [768768, 20482048], up to 4096*4096 for pure T2I). With image input the output ratio follows the last image |
n | integer | Normal mode: images to generate, 1-4, default 1. Set mode: maximum count 1-12, default 12 (the model decides the actual count) |
enable_sequential | boolean | Image-set mode, default false |
thinking_mode | boolean | Reasoning mode, default true (only in normal mode with no image input; better quality, slower) |
watermark | boolean | "AI-generated" stamp, default false |
seed | integer | [0, 2147483647] |
Vendor-native parameters (e.g. bbox_list box editing, color_palette custom palettes) go into parameters and pass through verbatim; whether they apply depends on the model you call.
Response
Same shape as Qwen image: output.choices[].message.content[].image carries the image URL, usage.output_image_count the actual count.
Billing
Billed by successfully generated image; the resolution tier affects the rate; failed requests are not billed. Rates on the Models page.
Error handling
Parameter errors return 4xx + {"code": "...", "message": "..."}. General error codes: Errors.