多元探索 API多元探索 API
Qwen Image

Qwen Image Image Editing API

Use `POST /v1/images/edits` to call `qwen-image-3.0` and `qwen-image-3.0-pro` for image-to-image/editing, supporting JSON multimodal requests and multipart file uploads.

POSThttps://zx1.deepwl.net/v1/images/edits
Request
curl --request POST 'https://zx1.deepwl.net/v1/images/edits' \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "qwen-image-3.0-pro",
    "input": {
      "messages": [
        {
          "role": "user",
          "content": [
            {
              "image": "https://example.com/reference.png"
            },
            {
              "text": "Keep the facial features, change the outfit to a modern professional style"
            }
          ]
        }
      ]
    },
    "parameters": {
      "prompt_extend": true,
      "size": "1080*1080",
      "n": 1,
      "watermark": false
    }
  }'
Response
{
  "created": 1788188983,
  "data": [
    {
      "url": "https://example.com/generated.png"
    }
  ],
  "metadata": {
    "output": {
      "choices": [
        {
          "finish_reason": "stop",
          "message": {
            "role": "assistant",
            "content": [
              {
                "type": "image",
                "image": "https://example.com/generated.png"
              }
            ]
          }
        }
      ]
    },
    "request_id": "example-request-id",
    "usage": {
      "input_image_count": 1,
      "input_image_type": "qima_input_1k",
      "output_image_count": 1,
      "output_image_type": "qima_output_1k",
      "output_width": 1080,
      "output_height": 1080
    }
  }
}

Image editing shares the model family with text-to-image, using the editing route. Two submission methods are supported: JSON multimodal requests and multipart/form-data file uploads.

Request Headers

Authorizationstring必填

Request authentication. Uses a Bearer token, e.g. Bearer YOUR_API_KEY.

Content-Typestring

Request content type. Must be application/json for JSON request bodies; for multipart/form-data it is carried automatically by the client (e.g. when curl uses -F), no need to set it manually.

Request Body

modelstring必填

Must be qwen-image-3.0 or qwen-image-3.0-pro.

inputobject

The native multimodal input structure containing messages.

messagesarray必填

Message list; pass only one message.

rolestring必填

Message role, fixed to user.

contentarray必填

Message content list. Each content element is either text or image. Image-to-image/editing passes 1–3 image items and one text.

textstring

Text content, i.e. the edit instruction.

imagestring

Reference image address, supporting HTTP/HTTPS URLs or Data URLs with a MIME prefix. Supported formats: JPG, JPEG, PNG, BMP, TIFF, WEBP, GIF; officially recommended width and height between 384 and 2048 pixels; a single image must not exceed 10 MB.

parametersobject

Qwen image generation parameters. When explicitly provided, the adapter uses these parameters directly and no longer copies size, n, or watermark from the top level.

sizestring

Output size in the native format WIDTH*HEIGHT (e.g. 1080*1080).

ninteger

Number of output images; the current upstream range is 1–6. When omitted or 0, it is treated as 1.

watermarkboolean

Whether to add a watermark.

prompt_extendboolean

Whether to enable prompt extension.

seedinteger

Random seed.

promptstring

Edit instruction used in the multipart format.

imagefile | string

Reference image file in the multipart format; can be repeated multiple times.

ninteger

Number of images in the multipart format. When omitted or 0, it is treated as 1.

watermarkboolean

Watermark switch in the multipart format.

multipart file uploads

Multipart requests submit the following fields with --form (see the Request Example panel on the right for a runnable example):

  • model
  • prompt
  • one or more image
  • n
  • watermark

Multiple files can be passed by repeating the image field:

--form 'image=@/absolute/path/reference-1.png' \
--form 'image=@/absolute/path/reference-2.png'

The multipart path does not build the full Qwen image parameters. When size, prompt_extend, or seed needs to be controlled, use the JSON multimodal request.

Request Example

Response Example

Response Fields

Image-to-image/editing shares the same response structure as text-to-image. The field definitions for created, data (with url, revised_prompt), metadata (with output, request_id, usage), and the error object (message, type, param, code), along with error-code meanings and troubleshooting, are documented in Image Generation API · Response Fields.

The only difference from text-to-image: metadata.usage.input_image_count reflects the number of reference images actually passed (0 for text-to-image).