Short answer: Alibaba Cloud Model Studio documents qwen-image-3.0 and qwen-image-3.0-pro for both text-to-image generation and image-to-image editing. For generation, describe the subject, composition, and visual direction. For editing, provide an image and state what to change. The examples below are editorial templates—not official or tested prompts. (help.aliyun.com)
Model names and API availability depend on the provider. The model IDs and endpoint below are specific to the documented Model Studio API; do not assume they apply to Qwen Chat or other hosts. (Model Studio API reference)
Choose a workflow by task
| Your task | Documented Model Studio option | What to know |
|---|---|---|
| Generate an image from text | qwen-image-3.0 or qwen-image-3.0-pro |
Model Studio lists both for text-to-image (T2I). |
| Edit an image using instructions and a reference image | qwen-image-3.0 or qwen-image-3.0-pro |
Model Studio lists both for image-to-image (I2I) as well as generation. For this route, omitting image means T2I; supplying image means I2I. (Model Studio API reference) |
| Try image editing in Qwen Chat | Qwen Chat’s “Image Editing” feature | Qwen’s article describes this chat experience. It does not establish which API or provider is available through the chat feature. (Qwen-Image-Edit article) |
Practical choice: If you need one documented Model Studio option for both generation and editing, evaluate the 3.0 models for your use case. The documentation cited here does not establish a universal winner or provide prices. Confirm the model and endpoint availability for your intended deployment before choosing. (Model Studio API reference)
A useful prompt structure
The sources cited here do not provide a single official prompt formula. As a practical starting point, make your request specific about:
- Subject and action: who or what is shown, and what it is doing.
- Composition: viewpoint, framing, placement, and background.
- Visual direction: medium, style, lighting, color, or mood.
- Text, if needed: give the exact words and where they should appear.
- Constraints: name important details to preserve or avoid.
These are original examples to adapt—not Qwen-authored templates, official sample prompts, or results from a model test.
Text-to-image example
Create a polished editorial illustration of a small neighborhood bakery on a rainy evening. Show the storefront at eye level, warm light glowing through the windows, and soft reflections on the sidewalk. Use a restrained palette of deep blue, amber, and cream, with clear negative space in the upper third for a headline. Do not include lettering in the image.
Workflow: Model Studio T2I with qwen-image-3.0 or qwen-image-3.0-pro. Input assumption: text prompt, with no image supplied. Intended result: a designed scene with headline space, rather than a finished image containing headline text. Model Studio documents T2I and the distinction between requests with and without an image input. (Model Studio API reference)
Local image-edit example
In the supplied photo, change the red mug on the table to dark green. Keep the table, hands, lighting, camera angle, and all other objects as close to the original as possible.
Workflow: Model Studio I2I with qwen-image-3.0 or qwen-image-3.0-pro. Input assumption: a source image plus the instruction. Intended result: one localized change. This is an editorial prompt example, not a documented guarantee that other pixels will remain unchanged.
As background for editing terminology, Qwen’s article about the separate Qwen-Image-Edit variant distinguishes localized “appearance” edits from broader “semantic” edits. That description is specific to the article’s subject; it does not establish that the 3.0 models behave identically. (Qwen-Image-Edit article)
Broader transformation example
Use the supplied portrait as the reference. Reimagine it as a hand-painted watercolor illustration, keeping the person recognizable while changing the overall visual style. Use a light paper texture and a muted spring palette.
Workflow: Model Studio I2I with qwen-image-3.0 or qwen-image-3.0-pro. Input assumption: a source portrait plus the instruction. Intended result: a broad stylistic change. This is an editorial example, not a claim about a particular model’s measured results. Qwen’s Qwen-Image-Edit article lists style transfer as an example of semantic editing, but that is background on that variant, not a 3.0 capability guarantee. (Qwen-Image-Edit article)
Text-in-image editing example
Replace the main headline on the supplied poster with “SPRING MARKET.” Keep it on the existing sign, centered, with lettering that fits the poster’s style. Do not change the illustrations or background.
Workflow: Alibaba Cloud Model Studio’s Qwen-Image-Edit workflow, using the documented DashScope example with qwen-image-3.0-pro. Input assumption: the poster as a source image plus the editing instruction. Intended result: a requested text replacement in a specified location. Model Studio documents that Qwen-Image-Edit can modify text in images, and its guide shows an API example using qwen-image-3.0-pro. This editorial prompt is not a tested result or a guarantee of exact rendering. This evidence applies to the documented Model Studio workflow; it does not establish identical behavior through Qwen Chat or other hosts. (Model Studio Qwen-Image-Edit guide) (help.aliyun.com)
Improve prompts without confusing the models
Qwen’s Qwen-Image-2.1 repository describes separate prompt-enhancer tasks: text-to-image prompt expansion and image-editing instruction rewriting. These are enhancer tasks, not the image-generation API workflow described below. The cited README excerpt does not establish additional technical details about enhancer inputs or architecture. (Qwen-Image-2.1 prompt enhancer README)
For a difficult local edit, you can also try refining in stages: request the main change, inspect the result, then describe any remaining correction. Qwen’s Qwen-Image-Edit article gives an example of a correction that did not work properly in one pass, followed by another localized refinement. This supports iterative refinement as an option, not a guarantee that repeated edits will succeed or that every API accepts marked regions. (Qwen-Image-Edit article)
Can you use Qwen image models through a production API?
Yes—Alibaba Cloud Model Studio documents a production API route for Qwen image generation and editing. Its reference lists qwen-image-3.0 and qwen-image-3.0-pro for T2I and I2I. The documented OpenAI-compatible endpoint includes a US (Virginia) route at /compatible-mode/v1/images/generations. For this route, the reference says that omitting image means T2I and supplying image means I2I. (Model Studio API reference)
The documented US endpoint has this form; substitute your workspace ID:
https://{WorkspaceId}.us-east-1.maas.aliyuncs.com/compatible-mode/v1/images/generations
Check the provider’s reference for the authentication setup and any access conditions that apply to your account and region. The endpoint and model IDs here describe Model Studio’s route only; they should not be generalized to other Qwen services. (Model Studio API reference)
Quick selection checklist
- Only generating images? Evaluate a documented T2I option such as
qwen-image-3.0orqwen-image-3.0-pro. - Editing an existing image? Use a documented I2I-capable option and supply the source image with a direct instruction.
- Changing one detail? State the target and the surrounding content you want preserved; inspect the output rather than assuming exact preservation.
- Editing words? Model Studio’s Qwen-Image-Edit guide documents text modification and demonstrates a DashScope workflow using
qwen-image-3.0-pro. Include the exact replacement text and its location, then proofread the rendered result. This is specific to the documented Model Studio workflow, not a guarantee for every Qwen host. (Model Studio Qwen-Image-Edit guide) (help.aliyun.com) - Using an API? Verify the provider’s model ID, endpoint, authentication setup, and input requirements. Model Studio documents its own route; do not generalize it to other Qwen services. (Model Studio API reference)
Sources
Alibaba Cloud Model Studio: Qwen-Image generation and editing API reference
Alibaba Cloud Model Studio: Qwen-Image-Edit guide
Qwen-Image-Edit: Image Editing with Higher Quality and Efficiency