Qwen & Wan Image Models | OpenModels Documentation

Qwen and Wan image models use the image generation endpoint with an input.messages payload. Use the same request shape for qwen-image-2.0, qwen-image-2.0-pro, wan2.7-image, and wan2.7-image-pro; choose model-specific options inside parameters.

curl --location --request POST "https://api.getopenmodels.com/v1/images" \
  --header "Authorization: Bearer $OM_API_KEY" \
  --header "Content-Type: application/json" \
  --data-raw '{
    "model": "qwen-image-2.0-pro",
    "route": { "provider": "<provider-for-this-model>" },
    "route_mode": "balanced",
    "input": {
      "messages": [
        {
          "role": "user",
          "content": [
            {
              "text": "Create a high-quality ecommerce poster for a white ceramic coffee cup, clean layout, realistic lighting, clear Chinese and English typography."
            }
          ]
        }
      ]
    },
    "parameters": {
      "n": 1,
      "watermark": false,
      "size": "2048*2048",
      "negative_prompt": "low resolution, blurry text, distorted hands, cluttered composition",
      "prompt_extend": true
    }
  }'

For Wan models, use the same payload shape and switch to Wan-specific parameter values:

curl --location --request POST "https://api.getopenmodels.com/v1/images" \
  --header "Authorization: Bearer $OM_API_KEY" \
  --header "Content-Type: application/json" \
  --data-raw '{
    "model": "wan2.7-image-pro",
    "route": { "provider": "<provider-for-this-model>" },
    "route_mode": "balanced",
    "input": {
      "messages": [
        {
          "role": "user",
          "content": [
            {
              "text": "A cinematic product photo of a premium black ceramic coffee cup on a marble table, soft side lighting, luxury ecommerce style."
            }
          ]
        }
      ]
    },
    "parameters": {
      "n": 1,
      "watermark": false,
      "size": "2K",
      "thinking_mode": true
    }
  }'

For Qwen and Wan image models, set output options inside parameters.

Field	Type	Notes
`model`	string	Supported values: `qwen-image-2.0`, `qwen-image-2.0-pro`, `wan2.7-image`, `wan2.7-image-pro`.
`input.messages`	array	Required. Use a single user message with `content` parts.
`input.messages[].content[].text`	string	Required for text-to-image prompts. Qwen 2.0 prompts support up to 1300 tokens; Wan prompts support up to 5000 characters.
`input.messages[].content[].image`	string	Optional image input for image editing or reference-image workflows when the selected model route supports image input. Use a public image URL or a base64 Data URL.
`parameters.n`	integer	Number of images. Qwen 2.0 currently requires `1`; Wan supports `1`-`4` when group generation is off, and `1`-`12` when `enable_sequential` is true.
`parameters.size`	string	Output size. Qwen 2.0 uses explicit pixel sizes such as `20482048`, `26881536`, or `15362688`. Wan supports `1K` and `2K`; `wan2.7-image-pro` also supports `4K` for text-to-image. Wan also accepts custom `WIDTHHEIGHT` values within its pixel limits.
`parameters.watermark`	boolean	Whether to add a watermark. Default: `false`.
`parameters.seed`	integer	Optional random seed in the range `[0, 2147483647]`. A fixed seed improves reproducibility but does not guarantee identical results.
`parameters.negative_prompt`	string	Qwen-only negative prompt, used to describe content to avoid.
`parameters.prompt_extend`	boolean	Qwen-only prompt optimization switch. Default: `true`.
`parameters.enable_sequential`	boolean	Wan-only group generation switch. Default: `false`.
`parameters.thinking_mode`	boolean	Wan-only quality mode. Default: `true`; only applies when group generation is off.
`parameters.color_palette`	array	Wan-only color theme control. Provide 3-10 color objects with `hex` and `ratio`; all ratios must sum to `100.00%`.