Qwen Image Edit
Calls Ali qwen-image-2.0-pro to perform single-image editing or multi-image fusion.
The public calling convention keeps top-level model, input, response_format, and exposes n, size, watermark, prompt_extend, negative_prompt, seed as top-level fields. Input images are placed in input.messages[0].content[].image, and the editing instruction is in a text item within the same array. The request body does not accept a top-level parameters field.
Authorizations
Pass Authorization: Bearer <token> in the request header.
Body
Image editing model name.
qwen-image-2.0-pro Unified response format.
url, b64_json Number of images to generate. The gateway maps this to upstream parameters.n.
1 <= x <= 6Output image resolution in widthxheight or width*height format. The gateway converts and maps this to upstream parameters.size.
Whether to add a watermark. The gateway maps this to upstream parameters.watermark.
Whether to enable smart prompt rewriting. The gateway maps this to upstream parameters.prompt_extend.
Negative prompt. The gateway maps this to upstream parameters.negative_prompt.
Random seed. Explicitly passing 0 is preserved and mapped to upstream parameters.seed.