Edit and recombine up to 10 images with Qwen 2.1 Image to Image. Optional inpainting mask for precise local edits. Try free — pay per generation. Qwen 2.1 Image to Image edits and recombines up to 10 reference images from a text prompt, with an optional inpainting mask for precise local edits.
Qwen 2.1 Image to Image edits and recombines up to 10 reference images from a text prompt, with an optional inpainting mask for precise local edits.
关于此模型
Qwen 2.1 Image to Image edits, restyles, and recombines up to 10 reference images from a single text prompt, preserving identity and composition where you want it kept. An optional inpainting mask switches the request into local-edit mode, changing only the white-marked region of one reference image while leaving the rest untouched. For pure text-to-image generation, see Qwen 2.1 Text to Image; for Alibaba's newer generation, see Qwen 3.0 Image to Image.
成本分析
| 提供商 | 费用 | 备注 |
|---|---|---|
| muapiapp | $0.03 (1K) / $0.05 (2K) | Flat per-image 定价 regardless of how many reference images are supplied. |
| Fal.ai | 暂不可用 | Qwen 2.1 is not currently listed on Fal.ai. |
| Replicate | 暂不可用 | Qwen 2.1 is not currently listed on Replicate. |
Flat per-image 定价 regardless of how many reference images are supplied.
Qwen 2.1 is not currently listed on Fal.ai.
Qwen 2.1 is not currently listed on Replicate.
** 竞品价格根据相似模型架构和使用层级估算。
配置参数
| 参数 | 类型 | 描述 | 默认值 |
|---|---|---|---|
| 提示词 | string | The prompt describing the result you want. When mask_url is supplied, describe what should be generated inside the white area of the mask. | Replace the backpack in the first image with the colour scheme from the second image, and leave everything else unchanged |
| 输入图像 | array | Reference images, given as an array of 1 to 10 URLs. Order matters when the prompt refers to 'the first image', 'the second image'. | https://d3adwkbyhxyrtq.cloudfront.net/ai-images/186/712345784292/4a8c5c70-abcc-4920-873e-b0e219986453.jpg |
| 蒙版 URL | string | Inpainting mask (white = area to change, black = kept unchanged) that switches the request into local-edit mode. Must be used with exactly one reference image. | - |
| 画面比例 | 枚举（10 个选项） | Output aspect ratio. Defaults to the first reference image's ratio. Ignored in local-edit mode. | auto |
| 分辨率 | 枚举（2 个选项） | Output resolution tier. 1K is the faster option; 2K produces four times the pixels and takes longer. | 1K |
| Background | 枚举（2 个选项） | Background type. Transparent outputs a PNG or WebP with a real alpha channel and cannot be used with JPEG. | opaque |
| 输出格式 | 枚举（3 个选项） | Output image format. jpeg has no alpha channel and cannot be combined with a transparent background. | png |
| Enhance Prompt | boolean | Rewrites the prompt into a fuller scene description before generating, which noticeably improves quality. Ignored in local-edit mode. | true |
| 种子 | integer | The random seed. Omit it to generate with a random seed. | - |
The prompt describing the result you want. When mask_url is supplied, describe what should be generated inside the white area of the mask.
Replace the backpack in the first image with the colour scheme from the second image, and leave everything else unchangedReference images, given as an array of 1 to 10 URLs. Order matters when the prompt refers to 'the first image', 'the second image'.
https://d3adwkbyhxyrtq.cloudfront.net/ai-images/186/712345784292/4a8c5c70-abcc-4920-873e-b0e219986453.jpgInpainting mask (white = area to change, black = kept unchanged) that switches the request into local-edit mode. Must be used with exactly one reference image.
-Output aspect ratio. Defaults to the first reference image's ratio. Ignored in local-edit mode.
autoOutput resolution tier. 1K is the faster option; 2K produces four times the pixels and takes longer.
1KBackground type. Transparent outputs a PNG or WebP with a real alpha channel and cannot be used with JPEG.
opaqueOutput image format. jpeg has no alpha channel and cannot be combined with a transparent background.
pngRewrites the prompt into a fuller scene description before generating, which noticeably improves quality. Ignored in local-edit mode.
trueThe random seed. Omit it to generate with a random seed.
-开发者文档
images_list; order matters when your prompt says "the first image", "the second image", etc.prompt describing the transformation you want.mask_url (white = change, black = keep) together with exactly one reference image to edit only that masked region.curl -X POST "https://api.muapi.ai/api/v1/qwen2-1-image-to-image" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"images_list": ["https://example.com/a.jpg", "https://example.com/b.jpg"],
"prompt": "Replace the backpack in the first image with the colour scheme from the second image, and leave everything else unchanged",
"resolution": "1K"
}'
request_id against the standard predictions result endpoint until status is completed.常见问答
Qwen 2.1 Image to Image is Alibaba's reference-image editing model — it edits, restyles, and recombines existing images from a text prompt instead of generating from scratch.
Between 1 and 10 images per request. More references reduce how much detail each individual one keeps, so stay within 4 when fidelity matters.
Supply `mask_url` — a black-and-white image where white marks the area to change — together with exactly one reference image. Only that white-marked region is regenerated; everything else is kept unchanged. Local-edit mode ignores `aspect_ratio` and `enhance_prompt`.
By default (`auto`) it matches the first reference image's aspect ratio, snapped to the nearest supported ratio. You can override it explicitly outside of local-edit mode.
Yes — set `background` to `transparent` with `png` or `webp` as the output format; it cannot be combined with a mask or with `jpeg`.