To edit images in Qwen Chat, open the official service, use its image-editing workflow, attach your existing photo, and describe the change you want. Specify what should stay unchanged, inspect the returned image, and correct any unwanted changes before saving it. Qwen’s official image-editing introduction documents edits such as changing colors, removing objects, replacing backgrounds, and modifying text within pictures.
Qwen’s public product page now calls the assistant Qwen Studio, although Qwen Chat remains the name used in many guides and official links. This article covers the official browser service—not a third-party image editor using a Qwen model.
Documentation reviewed: October 5, 2026. This guide is based on official documentation, not a completed signed-in editing test. Current menu layouts and export controls could not be independently confirmed. The prompts below are illustrative examples, not recorded Qwen results.
Before you upload a photo
Start with a clear image and one change you can judge easily. For a first attempt, changing an object’s color is more straightforward to evaluate than changing a person’s clothing, pose, face, lighting, and background together.
- Keep the original. Work from a copy and save useful edited versions separately.
- Remove unnecessary sensitive information. Check for visible addresses, account details, documents, or people who have not agreed to the upload.
- Define success before editing. Decide exactly what should change and which details would make the result unacceptable if altered.
The walkthrough uses an example photo of a blue matte ceramic mug with a white interior on a table. The task is to turn only its exterior and handle green. Use your own suitable photo and adjust the description to match it; do not ask the editor to preserve features that are not actually present.
How to edit images in Qwen Chat
1. Open the official Qwen service
Open Qwen Chat and complete any sign-in requirement shown before using the image tool. You do not need to follow a Python tutorial or configure a developer API key for the documented browser workflow.
Check the website you are using before uploading anything. An editor advertising “Qwen Image” may be a separate provider with its own account, charges, privacy policy, and controls. Our Qwen Studio guide explains the official assistant’s broader interface. Try-Qwen-AI.com is an independent guide, not the official Qwen service.
2. Use image editing, not just image analysis
Qwen’s published image-editing instructions direct users to an Image Editing feature. Select that option when it appears in your account. If your interface instead offers a combined image workflow, use its path for working from an existing image rather than requesting a text-only description.
Qwen also documents models that combine generation and editing. That does not establish the exact model or menu available in every account, so do not assume an older tutorial’s button layout must match yours.
The distinction is the output: an editing task must return an actual image you can inspect. A description of your photo, editing suggestions, or a message saying the task is complete is not an edited file. For creating a new picture without an existing source, use our separate Qwen image-generation guide.
3. Attach the correct source image
Use the image attachment control available in that workflow to add your source photo. Before submitting the edit, confirm that the correct image is attached and that no upload error remains.
For the mug exercise, attach only the original mug photo. Starting with one source avoids ambiguity about which image should be edited. Leave extra references out until a task genuinely needs them.
If the upload is rejected, follow the format or size requirement shown by the interface. Do not copy an API’s limits into the browser workflow. Converting a file to a supported format requires an actual image conversion; renaming its extension is not a substitute.
4. Describe the change and the details to preserve
A useful editing prompt identifies the source, target, change, protected details, and visual integration. “Make this better” leaves those decisions unspecified. “Change the mug’s exterior to green while keeping its shape and white interior” gives you a result you can check.
This reusable structure keeps the request focused:
Use the attached image as the base. Change [specific object or area] to [desired result]. Preserve [important details that must remain]. Match [relevant lighting, perspective, texture, or shadows]. Return the edited image.
For the example mug photo, the complete prompt is:
Edit the attached mug photo. Change only the blue exterior and handle of the mug to forest green. Keep the white interior, mug shape, handle position, table, background, camera angle, and framing unchanged. Preserve the matte ceramic finish and the original lighting direction. Do not add text, decorations, or other objects. Return the edited image, not a description of the changes.
Submit the request with the photo attached. These preservation instructions describe your requirements; they do not guarantee that every other pixel will remain identical.
5. Inspect the result before accepting it
Compare the returned image with your original at a useful viewing size. Do not judge the edit only from a small thumbnail or the assistant’s written explanation.
- Requested change: Is the exterior green, including the handle, without green spreading into the white interior?
- Protected details: Are the mug’s outline, handle position, table, and framing still correct?
- Integration: Do the finish, highlights, contact shadow, and edges look consistent with the scene?
- Unwanted additions: Has the edit introduced a logo, pattern, extra object, or distorted background detail?
For this exercise, a green mug is not sufficient if the handle has changed shape or the interior has also been recolored. Accept the result against the criteria you set, not just its overall attractiveness.
6. Correct one remaining problem at a time
Suppose the color and shape are right, but the mug now looks glossy. That is a hypothetical follow-up—not a reported test result. Select or reattach the green version you want to refine, then identify that version explicitly:
Use the attached green-mug image as the base. Change only the mug's glossy finish to a soft matte ceramic finish. Reduce the sharp, mirror-like highlights without changing the green color. Keep the mug shape, white interior, handle, table, background, camera angle, and framing as they are.
Avoid “fix the previous one” when several images are present. Make the intended source unambiguous, then check the new output again.
If the result has drifted substantially from the original, return to the original or your best saved version rather than repeatedly editing an increasingly inaccurate image. Even Qwen’s consistency improvements are described as reducing image drift, not guaranteeing identical preservation.
7. Save and check the actual image file
Use the output’s download option if your interface provides one. The current Qwen export controls were not independently confirmed for this guide, so do not assume a particular icon, position, or file-format menu.
When the returned picture is exposed as a normal browser image, Chrome also supports right-clicking it and choosing Save Image As, as documented in Google’s download instructions. A small embedded preview may not be the largest available version; inspect the saved file rather than assuming it is full resolution.
Open the saved file locally and check its dimensions, crop, and visible quality. Keep the original and useful intermediate versions. A screenshot of the chat is useful for documenting the process, but it is not a substitute for the edited image file.
Qwen image-editing prompts for common tasks
The following examples apply the same change-and-preserve approach to other tasks. Adapt object names and positions to your actual photo, and test each request separately rather than combining the examples into one prompt.
Replace a product background
Replace the background behind the attached product with a plain, light-gray studio backdrop. Keep the product's shape, proportions, color, label, and camera angle. Retain a natural contact shadow so the product does not appear to float. Do not rewrite the label or add props. Keep the original framing.
Review the product silhouette and every visible label character. An attractive background is not a usable product photo if the packaging, color, or branding has changed.
Remove an unwanted object
Remove the red suitcase at the lower-left edge of the attached photo. Fill that area with a plausible continuation of the visible pavement, matching its perspective, texture, and lighting. Keep the people, buildings, remaining objects, and framing unchanged.
Look for repeated textures, distorted edges, or an object shadow left behind. Treat the reconstructed area as a plausible replacement, not evidence of what was actually hidden behind the removed object.
Change text inside an image
Qwen’s official editing examples include changing English and Chinese text within pictures. Give the old wording, new wording, and location rather than asking for a general text correction.
In the attached poster, replace the headline "SUMMER SALE" at the top center with "AUTUMN SALE". Keep the other text and graphics unchanged. Match the existing headline's font style, alignment, perspective, and visual weight as closely as possible. Use exactly "AUTUMN SALE" with no additional words.
Proofread the downloaded result character by character, including numbers and small text outside the edited area. A visually similar font is not proof of an exact match. When typography or legal wording must be exact, make the final text change in an editor with an editable text layer.
Improve a portrait without requesting a different identity
Adjust only the exposure and white balance of the attached portrait for a natural, evenly lit appearance. Preserve the person's facial features, skin texture, expression, hairline, hairstyle, clothing, pose, and background. Do not reshape the face, change apparent age, or add beauty filters.
Compare facial landmarks, hair, skin texture, and expression against the original. “Preserve the face” is an instruction to evaluate, not a guarantee of identity preservation. For small exposure corrections where every feature must remain untouched, a conventional adjustment may be more appropriate than a generative edit.
Restore a damaged photo cautiously
Reduce the visible scratches and dust in the attached old photograph. Keep the faces, clothing, objects, composition, and existing tonal style. Avoid inventing fine facial details or adding objects. Do not colorize the image or change its historical appearance.
Compare repaired areas with the source and keep the untouched scan. Use the result as a restoration interpretation, not proof that newly reconstructed details were present in the original scene.
Using more than one reference image
Qwen’s published model examples include edits guided by multiple images. That is not a universal upload allowance for every Chat account. Use multiple references only when the image workflow you are using accepts them.
Assign each image a role. The base image supplies the composition; another image may supply a material or color reference. Do not leave Qwen to guess which photo should be transformed.
Image 1 is the room photo to edit. Image 2 is a fabric reference only. In Image 1, change only the sofa upholstery to resemble the fabric texture and color in Image 2. Keep the sofa's shape, position, and size, along with the room layout, windows, other furniture, camera angle, and lighting direction. Do not insert Image 2 itself into the room.
Confirm that the image numbering matches the attachments. After generation, compare against the base image, not just the reference: the fabric can look convincing while the sofa or room geometry has changed incorrectly.
When Qwen image editing is not working
Identify the stage that failed before changing settings or repeating the request. A missing feature, rejected upload, text-only answer, and unwanted visual change need different checks.
| What happens | What to check next |
|---|---|
| The image-editing option is missing. | Confirm that you are using the official service, complete any sign-in requirement, and read availability or compatibility notices. A missing button alone does not identify a regional restriction. |
| Qwen describes the image instead of editing it. | Check that the source is attached and that you are using an image-output workflow. State the concrete edit and request an edited image, not an explanation. |
| The image will not upload. | Read the actual upload error. Try one non-sensitive image that meets the interface’s stated requirements. Convert unsupported files properly rather than renaming them. |
| A limit or generation error appears. | Follow the displayed message. Distinguish an exhausted allowance, temporary failure, unsupported request, and content restriction. Do not assume they share a reset time or fix. |
| The wrong object or source image changes. | Name the object and its location, clarify the role of each attachment, or reattach only the intended base image. |
| The edit damages faces, labels, or fine detail. | Return to a better source version, narrow the request, and inspect the affected areas again. Use a controlled editor for details that must remain exact. |
A message about an unsupported system is not the same as a country restriction or an exhausted image quota. Follow the stated issue instead of assuming that a VPN, new account, or paid plan will fix it.
For more detailed distinctions between browser tools, file requirements, and export problems, see our Qwen image limits, formats, and troubleshooting guide.
Is editing images in Qwen Chat free?
Qwen’s public product page describes Qwen Studio as free to use. That statement does not establish unlimited image edits, a fixed daily allowance, or identical feature access for every account.
Use the allowance or restriction displayed in your own account. An API trial balance, a third-party editor’s credits, or a coding subscription does not establish your Chat image quota. Our Qwen pricing guide separates the consumer service from developer and other paid offerings.
Check transparency and other output requirements
Removing a background does not, by itself, prove that the downloaded file is transparent. A white background or a checkerboard drawn into the picture can look like a cutout without providing actual transparency.
To verify it, place the saved image over a colored layer in an editor. The surrounding area should reveal that layer rather than show white pixels or a fixed checkerboard. Do not promise transparent exports based only on a model announcement or an API specification.
Apply the same principle to dimensions, brand colors, and editable layers: check the delivered file against the requirement. A prompt asking for an exact color, layout, or size is not confirmation that the saved result meets it.
Protect private images and review publication rights
Before uploading confidential material, read the Qwen Privacy Policy. It states that de-identified user content and feedback may be used for model training and improvement. Do not assume that editing through a browser means the image stays on your device or is excluded from training.
Use images you are authorized to upload and edit, and review the service terms before commercial publication. Do not treat a downloadable model’s license as a complete answer to questions about the hosted service, a photographer’s rights, trademarks, or the people shown in a photo.
Before publishing, compare the saved edit with the original one final time. The useful result is not simply a more polished picture: it is a file that makes the intended change, preserves the details that matter, and is appropriate for its stated use.