Skip to main content
POST

Authorizations

string
required
All requests require Bearer Token authentication.Visit the API Key Management page to obtain your API Key, then add it to the request header:

Available Models

Billing is based on successfully generated images × unit price. Input is not billed. Resolution and aspect ratio do not affect pricing. Failed requests are not charged.

Body

string
required
Image generation model name.
  • wan2.7-image-pro — Professional edition, up to 4K for text-to-image
  • wan2.7-image — Standard edition, faster, up to 2K
string
Text description for image generation, up to 5000 characters.
  • Text-to-image (no image_urls): required
  • Image editing (with image_urls): optional but recommended
Example: "A flower shop with exquisite windows, beautiful wooden door, flowers on display"
array<string>
Input image URL array for editing and multi-image reference scenarios.Providing this field switches the request to image editing mode.Supported formats: HTTP/HTTPS URLs; data:image/...;base64,... Base64Constraints: Up to 9 images; JPEG / PNG / WEBP / BMP; 240–8000 px, aspect ratio 1:8 ~ 8:1; ≤ 20MB per image
Output aspect ratio automatically matches the last input image. Editing mode supports up to 2K only — 4K is not available.
integer
default:"1"
Number of images to generate.
  • Standard mode: 1–4 (default 1)
  • Sequential mode (enable_sequential: true): 1–12 (default 1)
Billed per successfully generated image. Pre-charged based on n.
string
Output resolution or aspect ratio. Supports three formats:① Resolution keyword (recommended): 1K / 2K (default) / 4K (wan2.7-image-pro text-to-image only)② Aspect ratio: 1:1 / 16:9 / 9:16 / 4:3 / 3:4 / 3:2 / 2:3 (defaults to 2K tier)③ Pixel dimensions: 1024x1024 or 1024*1024
string
Resolution tier keyword: 1K / 2K / 4K. Can be combined with size (aspect ratio).
string
Negative prompt describing elements to avoid. Example: "blurry, distorted, low quality"
boolean
default:"false"
Whether to add an “AI Generated” watermark to the bottom-right corner.
integer
Random seed, range 0–2147483647. Same seed with identical parameters produces visually consistent results.
boolean
default:"true"
Enable enhanced reasoning mode to improve image quality at the cost of longer generation time.
Only effective when sequential mode is disabled and no image input is provided.
boolean
default:"false"
Enable sequential image generation mode — generates multiple thematically coherent images in one request. Ideal for storyboards and series.
  • Maximum n is 12 when enabled
  • thinking_mode and color_palette are ignored in sequential mode
  • wan2.7-image-pro supports up to 2K in sequential mode (4K not supported)
array
Bounding boxes for interactive editing — specifies exact regions to edit or insert content.Structure: [[[x1, y1, x2, y2], ...], ...]
  • Outer array length must equal the length of image_urls
  • Pass [] for images with no bounding box
  • Max 2 boxes per image; coordinates are absolute pixel values, origin (0,0) at top-left
Example: [[], [[989, 515, 1138, 681]]]
array<object>
Custom color theme. Standard mode only (not sequential mode).
  • 3–10 entries (8 recommended); each entry requires hex and ratio
  • All ratio values must sum to exactly 100.00%

Response

string
Response status. Returns "success" on success.
array

Examples

Text-to-Image (minimal)

Text-to-Image (with resolution)

Text-to-Image (custom color palette)

Sequential Image Generation

Single Image Editing

Multi-Image Reference / Element Fusion

Interactive Editing (Bounding Box)

bbox_list corresponds 1-to-1 with image_urls. Pass [] for images with no selection.
Querying ResultsImage generation is asynchronous. Poll the Task Status endpoint using the returned task_id until status == completed.