Qwen Image 3.0 Pro Image to Image: Edit Photos by Prompt

Last verified: October 19, 2026

Qwen Image 3.0 Pro edits and restyles the photos you upload based on a written instruction. The prompt box here takes up to 5,000 characters, so you have room to describe every change, every element that has to stay, and how several source images fit together. Alibaba's Model Studio guide recommends starting image edits with this model, and it accepts a negative prompt for anything you want kept out of the result.

Each edit request sends your images together with a single text instruction. That means the prompt has to carry the whole job: which picture supplies the subject, which one supplies the outfit or product, what changes and what stays fixed. A short request leaves all of those decisions to the model. A full brief makes them for it.

On the form, Aspect Ratio and Resolution Quality set the frame and size of each result separately from the brief. One well-written prompt can therefore produce both a square post and a vertical story from the same source.

Explore more image to image

Image To Image Seededit V3 Nano Banana 2 GPT Image 2 Seedream 4.5 Edit Qwen Image 3.0 Recraft V3
Capability Snapshot

Qwen Image 3.0 Pro editing specs

What you can put in, the frames you can pick, and what you get back from each run.

Input

Up to 3 images (JPG, JPEG, WEBP or PNG, up to 10 MB each) + a text prompt of up to 5,000 characters

Resolution

1K or 2K

Aspect Ratio

1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9

Output

1 image per run

Techniques for more controlled edits

Give every upload a job in the prompt

When you add more than one image, the model numbers them by input order: the first is image 1, the second is image 2. It also understands bracketed markers such as [image 1]. Write out each image's role, for example identity from image 1, garment from image 2 and pose from image 3, and add a few words on what each picture shows. That way the instruction still points to the right file if the order isn't what you expected. Alibaba's own example shows why this matters: the same two images give opposite results for 'move image 1 onto image 2' and 'move image 2 onto image 1'.

Keep exclusions in Negative Prompt

Use the main prompt for the change itself and put anything that must not appear in the Negative Prompt field. Alibaba describes this parameter as the place for content to exclude, gives blur and extra fingers as examples, and says it helps improve output quality. For portrait edits, terms like different person or altered facial features target the change you least want. For product shots, try extra logos or warped label text. Keeping these out of the main box also leaves its full length for the brief.

Prepare sources within the recommended pixel range

Alibaba recommends input images between 384 and 2048 pixels for the best editing results. Before you upload, resize very large camera files down into that range. Avoid sources that only just clear the uploader's minimum, because a tiny file gives the model little detail to keep. Apply the same check to every reference image you add, not only the main photo.

How do you edit a photo with Qwen Image 3.0 Pro?

Five steps from source photo to finished edit.

Why did my edit come out wrong?

Common problems with uploads, identity, prompt length and variations, and how to fix each one.

My photo won't upload.

The file is probably outside what Add Image accepts. It may be in a format the uploader doesn't list (HEIC files from phones are a common case), it may be over the size limit, or it may be smaller than the minimum dimensions. The upload question in the FAQ gives the exact limits. Convert the file to one of the listed formats, export a smaller version if it's too large, and replace a tiny source with a larger original. Then add it again. You don't lose anything, because generation only starts when you click Generate.

The person's face looks different after the edit.

If the prompt only describes the change, nothing in it tells the model what to protect. Alibaba's own portrait example tells the model to preserve the woman's facial features and hair exactly before it describes her new outfit and setting. Do the same: name the face, hairstyle, skin tone and anything else that makes the person recognisable, then describe the change. Add identity-related terms to the Negative Prompt field too, and run it again with the same source.

My prompt won't fit in the box.

The Prompt box stops accepting text at its character limit. Start by cutting repeated adjectives and descriptions that say the same thing twice. Describe each image's role only once. Move everything you want kept out of the image into Negative Prompt, so the main box holds only the change and what must stay. Then paste in the shorter version and generate again.

I only got one image and wanted several options to choose from.

Each run returns a single image, so variations come from running again. Keep the sources and settings the same and change one thing at a time, such as a phrase in the prompt, the Aspect Ratio or the Resolution Quality. That way you can tell which change caused which difference. Then keep the run you like best.

Frequently Asked Questions

Which images can I upload, and what are the limits?

Add Image takes one to three images per run, and at least one is required. Accepted formats are JPG, JPEG, WEBP and PNG. Each file can be up to 10 MB and must be at least 300×300 pixels.

What kinds of changes can Qwen Image 3.0 Pro make to a photo?

Alibaba's editing guide, which covers the 3.0 series, lists these edits: changing text inside an image, adding, deleting or moving objects, changing what a subject is doing, transferring styles and enhancing details. A full restyle and a small fix, like replacing a sign's wording, both work the same way: describe the result you want and say what should stay the same.

Does Enhance Prompt rewrite my instruction automatically?

No. Enhance Prompt sits under the prompt box and suggests fuller prompts based on a rough idea. Your text only changes when you choose one of the suggestions. After you pick one, read it through and check that it still names what must stay unchanged and what each uploaded image is for.

What makes one generation cost more than another?

The credit cost of a run is calculated from the options you choose and shown above the Generate button once you're signed in. It changes with the Resolution Quality you select and the number of images you upload. Using Enhance Prompt doesn't use credits.

How do I fix one detail in a result without starting over?

Download the result, add it in Add Image as your new source, and write a short instruction that names only the detail to change and says everything else stays as it is. A narrow follow-up edit leaves the parts you already like alone, which is easier to control than rewriting the original brief.

References

Sources and citations used to support the content provided above.

Updated: 2026-10-19 23:37:18 6 Sources

www.alibabacloud.com

Source Link
https://www.alibabacloud.com/help/en/model-studio/image-model/

docs.modelstudio.console.alibabacloud.com

Source Link
https://docs.modelstudio.console.alibabacloud.com/en/model-studio/qwen-image-generation-and-editing-api-reference

help.aliyun.com

Source Link
https://help.aliyun.com/en/model-studio/qwen-image-generation-and-editing-api-reference

qwen.ai

Source Link
https://qwen.ai/blog?id=qwen-image-3.0

www.alibabacloud.com

Source Link
https://www.alibabacloud.com/blog/qwen-image-3-0-rich-content-authentic-details-deep-knowledge_603385

www.alibabacloud.com

Source Link
https://www.alibabacloud.com/help/en/model-studio/qwen-image-edit-guide