QwenWork adds custom image sizes: decide where an image will go before generating it

In its October 15 release of version 1.3.1, QwenWork added support for custom dimensions and quality levels in image generation and editing. Alongside what the image should depict, where it will be used can enter the process earlier.

The change is useful for people who repeatedly make banners, covers and event graphics. The release notes do not list every size supported by every model, or promise arbitrary dimensions. Follow the options and results actually available in your task.

Suppose you need a landscape graphic for a book club. Generating a square first and forcing it into a banner may cut off a person or the space intended for a title. Giving AI the intended use, dimensions and text area up front may help the composition fit its destination from the start.

Separate the request into three parts: content, such as books, a table and readers; layout, such as a scene on the left and empty space on the right; and delivery requirements, such as the dimensions required by the page. Feedback then becomes more specific: the scene is wrong, or the empty space is too small, rather than simply “make it look better.”

Quality options provide another production choice, but they do not replace inspection. A useful file fits the target page, keeps its subject clear and makes necessary text readable. A detailed-looking preview alone does not prove that it meets the requirements.

Update to a client that includes the 1.3.1 capability, sign in and confirm that your account can generate images. The official desktop guide describes starting from New task, continuing edits in the same conversation, and previewing or downloading files from the output card.

1. Define the destination first. Note the target page’s required width and height, then start an image request from New task. Treat dimensions as a delivery requirement to verify; “landscape” is not a precise size.
2. Describe the composition too. For example: “Make a bright landscape illustration for a book club. The target is 1536 × 1024 pixels; if that is unsupported, tell me first. Put the desk and readers on the left. Leave roughly the right third clear for an event title to be added later. Do not generate text in the image.” This is a suggested input to adapt to your material.
3. Confirm whether the requirement can be met. That size is an example request, not a verified supported option. The official notes do not explain the exact controls for size and quality. Check the feedback actually shown before and after the task; a prompt is not proof that a parameter took effect.
4. Put the result back in its intended setting. Download it and check the file dimensions, complete subject and empty space. Preview it in the target page to see what a narrow screen might crop.
5. Ask for a specific change. For example: “Keep these colors, move the reader further left, and add no objects on the right.” Check dimensions and layout again after saving the revision.

If the final graphic needs dates, an address or other important text, consider adding those in a layout tool and checking every word. A date change then does not require remaking the illustration.