The Qwen image generator is available through two official no-code routes: the Create Image experience in Qwen Chat and the dedicated Qwen Create workspace. Open the route available to your account, choose image generation rather than ordinary chat or video, describe the finished asset you need, select an available aspect ratio, generate a draft, and inspect the result before downloading it.
This independent guide focuses on creating a new image from text. It does not treat every website with “Qwen Image” in its domain as official, and it does not apply API specifications automatically to the consumer interface. For general account and interface guidance, see our Qwen Studio guide.
Which Qwen image-generation route should you use?
| Route | Best for | Important distinction |
|---|---|---|
| Qwen Chat | Beginners, prompt development, and quick no-code generations | The interface may choose or simplify the underlying image model. |
| Qwen Create | A focused creative workspace with AI Image, AI Video, and Assets areas | Availability and controls can vary by region, account, or device. |
| Alibaba Cloud Model Studio API | Automation, named models, repeatable parameters, and production workflows | It uses separate credentials, regional endpoints, billing, and limits. |
| Downloadable Qwen image model | Technical local research, evaluation, and custom infrastructure | The model license and hardware requirements are separate from consumer-service terms. |
The model name shown at the top of a general Qwen conversation does not necessarily identify the model that renders an image. Current Model Studio documentation lists developer Model IDs including qwen-image-3.0-pro, qwen-image-3.0, and qwen-image-2.1-pro. The consumer interfaces may hide those IDs or offer different labels.
The downloadable Qwen-Image-2.1 model is another distinct route. Its official model card describes unified generation and editing, transparent-image support, and multi-reference editing. Do not assume that its local controls, license, or hardware requirements are the same as Qwen Chat, Qwen Create, or the hosted API.
How to generate an image in Qwen Chat
- Open the official image entry. Go to Qwen Chat Create Image and sign in if the service asks you to.
- Confirm that image generation is selected. Choose the displayed Create Image or equivalent tool rather than sending the request to an ordinary text-only conversation.
- Start a clean task. A new conversation reduces the chance that unrelated instructions from an earlier chat affect the visual brief.
- Name the finished asset. State whether you need a website hero, social post, poster, thumbnail, product concept, illustration, or presentation visual.
- Choose the canvas before generating. Select an available ratio or size that matches the publishing destination.
- Enter a structured prompt. Describe the subject, composition, environment, style, lighting, exact visible text, and constraints.
- Generate a draft and inspect it at delivery size. Check text, object count, anatomy, edges, reflections, crop safety, and factual details.
- Revise only the failed layer. Keep the successful subject, framing, style, or lighting instead of rewriting the entire brief after one defect.
- Download the selected image. Save the prompt, date, route, displayed model or mode when known, ratio, and any later manual edits with the exported file.
Use Qwen Create when the dedicated workspace is available
Qwen Create is the more focused visual route. Its public interface identifies separate AI Image, AI Video, and Assets areas. Choose AI Image for a new still image, then use the model, quality, ratio, or asset controls actually displayed in your account.
If Qwen Create displays a region, account, device, or compatibility notice, use another officially available Qwen route rather than attempting to bypass the restriction. The screenshot below records one access result from the existing page; it does not prove worldwide unavailability.
Plan the image before writing the prompt
A useful prompt begins with the deliverable, not a pile of quality adjectives. Make these decisions before the first generation:
- Destination: website hero, YouTube thumbnail, social post, poster, slide, product card, or print draft.
- Mandatory subject details: object count, clothing, product shape, materials, colors, and action.
- Composition: subject placement, camera angle, foreground, background, copy space, margins, and reading order.
- Visual language: photograph, editorial illustration, 3D render, vector design, watercolor, or another single dominant style.
- Exact copy: every word, number, line break, language, and capitalization that must appear.
- Constraints: elements that must not appear and details that must remain unchanged.
An aspect ratio describes the shape of the canvas; resolution describes its pixel dimensions. Writing “8K” in a prompt does not force an 8K file. Use the size or quality controls actually offered by the selected route.
| Ratio | Typical destination | Composition advice |
|---|---|---|
| 16:9 | Website heroes, YouTube thumbnails, and presentations | Use horizontal flow and reserve copy space deliberately. |
| 9:16 | Stories, Reels, Shorts, and vertical posters | Keep important content away from top and bottom interface overlays. |
| 1:1 | Square posts, cards, and profile grids | Use a centered or balanced layout with limited empty side space. |
| 4:5 | Portrait social posts and advertising creatives | Leave enough vertical room for the subject and supporting copy. |
How to write a better Qwen image prompt
Write the request as a compact production brief. Put mandatory requirements before decorative language, and separate exact visible text from the visual description.
Create [finished deliverable] for [destination or audience].
Subject: [main subject, quantity, materials, clothing, or defining details].
Scene: [environment, time, and context].
Composition: [placement, scale, foreground, background, margins, and copy space].
Viewpoint: [shot size, camera angle, or perspective].
Lighting and color: [direction, softness, palette, and mood].
Visual style: [one dominant style].
Exact visible text: “[copy]” or “No text.”
Constraints: [three to five elements to preserve or exclude].
Output: [aspect ratio and orientation].Documented example: a website hero image
The following prompt was used in the existing page’s Qwen Studio example on September 8, 2026. It demonstrates a practical way to control subject placement and reserve usable copy space. The screenshots document that specific session; they do not guarantee an identical result from another account or later model.
Create a landscape website hero image for a sustainable architecture company.
Subject: a compact solar-powered wooden cabin with realistic materials and restrained modern design.
Scene: a forested hillside at sunrise with light morning mist.
Composition: position the cabin in the right third of the frame and leave clean, low-detail copy space on the left.
Viewpoint: eye-level wide architectural photograph with natural perspective.
Lighting: warm sunrise light from the upper left, soft shadows, realistic reflections.
Visual style: premium editorial architectural photography.
Exact visible text: no text.
Constraints: no people, no logos, no floating objects, no exaggerated glass walls, no fantasy structures.
Output: 16:9 landscape composition.
How to create posters and images with text
The official Qwen-Image-3.0 announcement and current developer documentation emphasize complex layouts and multilingual text rendering. Treat this as a capability claim, not a promise that every character will be correct in every consumer generation.
Place required copy in its own section, state the language and line breaks, define the hierarchy, and divide the canvas into clear zones. Keep the amount of generated text short when accuracy matters.
Create a 4:5 portrait event poster with wide outer margins and an off-white background.
EXACT VISIBLE TEXT
Small category: “PRODUCT DESIGN WORKSHOP”
Main headline on two lines:
“Design with
Clarity”
Subheading: “A practical workshop for product teams”
Date: “September 18”
Button label: “Register Now”
LAYOUT
Top 35%: category, headline, and subheading, left-aligned.
Middle 45%: one elegant abstract illustration made from translucent geometric layers.
Bottom 20%: date on the left and a rounded call-to-action on the right.
STYLE
Modern editorial layout, dark navy typography, restrained violet accents, and generous spacing.
CONSTRAINTS
Use the exact wording and capitalization. Do not translate, repeat, add, or misspell text. No logos, QR codes, extra icons, or decorative words.This poster prompt is a suggested example, not a newly executed test. Review every word, number, punctuation mark, line break, and layout relationship. For contracts, prices, medical instructions, legal notices, ingredients, or other mission-critical copy, generate the visual concept first and add the final verified text in a conventional design editor.
Review the image before downloading it
Inspect the result as a deliverable at its intended publishing size. A polished preview can still contain wrong text, duplicated objects, impossible reflections, or inaccurate product details.
| Check | What to inspect | Pass condition |
|---|---|---|
| Prompt adherence | Subject, action, environment, object count, and exclusions | Every mandatory requirement is present and no prohibited element changes the meaning. |
| Composition | Placement, margins, copy space, crop, and hierarchy | The layout works at the final ratio and viewing size. |
| Text | Spelling, capitalization, punctuation, numbers, and line breaks | Every character is correct or will be replaced before publication. |
| People and objects | Hands, eyes, limbs, edges, reflections, shadows, and duplicates | No distracting structural error remains. |
| Product fidelity | Shape, materials, controls, packaging, and labels | The image does not misrepresent an actual product. |
| Facts | Charts, diagrams, maps, labels, prices, and educational claims | Every factual element matches an authoritative source. |
| Rights and privacy | Faces, logos, copyrighted material, private data, and consent | The required permissions and publishing basis are documented. |
Revise one failing layer at a time
- If the subject is too small, change the framing or percentage of the canvas without replacing the style.
- If the headline is wrong, preserve the scene and rewrite only the exact-text and layout instructions.
- If extra objects appear, state the exact count and remove unnecessary background detail.
- If the layout is crowded, reduce the number of elements and define named zones and margins.
- If an actual product changes shape, use a supported reference-image or editing workflow rather than describing it repeatedly from memory.
Creating a new image and editing an existing one are different workflows. For background replacement, object removal, text changes, or reference-guided edits, use our separate Qwen image-editing guide.
Common Qwen image generator problems
| Problem | What to check next |
|---|---|
| The image tool is missing | Confirm that you are using official Qwen Chat or Qwen Create, sign in if required, and check the account, device, and availability notice. |
| Qwen answers with text instead of an image | Select Create Image or the equivalent image tool before sending the prompt. |
| The result has the wrong shape | Choose the ratio before generation and state the destination and orientation in the prompt. |
| Visible text is misspelled | Shorten the copy, isolate it from visual instructions, define line breaks, and replace critical text in a design editor. |
| The subject is distorted or duplicated | Simplify the scene, specify the exact count, enlarge the main subject, or use a reference-guided edit. |
| Qwen Create is unavailable | Use an officially available route such as Qwen Chat and do not attempt to bypass a geographic, account, or platform restriction. |
| Generation or download fails | Read the displayed error, check limits and supported settings, avoid repeated duplicate jobs, and retry later when the notice indicates a temporary problem. |
For a deeper separation of consumer-interface limits, API formats, temporary result links, and recurring errors, see Qwen Create limits, formats, exports, and errors.
Is the Qwen image generator free?
Qwen Chat or Qwen Create may allow image generation without a separate per-image payment in a particular account, region, promotion, or session. That does not establish permanently free or unlimited use. Sign-in requirements, queues, daily allowances, quality modes, and access notices can change.
Hosted API use has separate pricing and free-quota rules, and a third-party website sets its own credits and subscriptions. Check the notice shown in the exact service you are using and review our Qwen pricing guide before assuming that consumer access includes API automation or unrestricted production use.
When to move beyond the no-code generator
- Edit an existing image: use an official editing route when you need to preserve a face, product, layout, or background while changing a specific element.
- Automate generation: use the official Model Studio image API when you need named models, regional endpoints, seeds, negative prompts, repeatable sizes, multiple outputs, and application integration. Those specifications do not define Qwen Chat limits.
- Run a model locally: evaluate the hardware, software, model card, and license for the exact downloadable checkpoint. Qwen-Image-2.1 is distributed under the Qwen Research License, whose standard grant is limited to non-commercial purposes unless a separate commercial license is obtained.
- Use a third-party host: verify its identity, exact model version, price, data policy, output terms, and support. A domain containing “Qwen Image” is not automatically operated by Qwen.
Developers can use the site’s Qwen API guides for broader setup and security guidance. Do not automate a consumer interface by extracting browser tokens or reusing session cookies.
Commercial use, ownership, and privacy
Commercial use depends on the route you choose, its current terms, the model license where applicable, your source materials, and the intended jurisdiction and publication context. Review the official Qwen Terms of Service and Usage Policy for the consumer service, or the governing cloud agreement for an API deployment.
Do not assume that a generated image is unique, copyrightable, trademark-safe, non-infringing, factually correct, or cleared for every commercial campaign. Check recognizable people, logos, artwork, characters, branded packaging, product claims, and required AI-content disclosures before publication.
Before uploading a reference image, review the policy of the exact service. Remove private metadata and identifiers, obtain permission for faces and copyrighted assets, and avoid uploading API keys, identity documents, confidential client work, customer data, unreleased products, or personal photographs without authorization. The Qwen Privacy Policy applies to the Qwen service it describes; an API provider or third-party host may use different terms. Our Qwen privacy guide explains the questions to check.
Last verified: October 5, 2026. Official Qwen product pages, the current Model Studio image API reference, the Qwen-Image-2.1 model card and license, and the live owner page were reviewed. Existing September 2026 Qwen Studio screenshots and the earlier Qwen Create access screenshot were retained. No new authenticated image generation, paid API call, purchase, or local-model test was performed for this revision; the poster prompt is a suggested example rather than a recorded result.