Qwen Image 3.0 takes no --flags — it is not Midjourney. You brief it like a creative director in plain language, and it renders text better than almost anything else. This cheat sheet is the one-page reference: the brief formula, then copy-paste tables for ratios, text, lighting, lens and style, plus five ready prompts. For deeper reads, see the full best Qwen Image prompts roundup and the photorealism guide.

Advertisement

The prompt formula

A good Qwen prompt reads like a creative brief, not a keyword salad. Cover six things in order — subject and audience, composition, action, setting, lighting, and style — then close with a one-line constraints statement.

[Subject + audience]: who or what this is for and the main subject.
[Composition]: framing, angle, where elements sit, negative space.
[Action / pose]: what the subject is doing.
[Setting]: location, background, props, mood.
[Lighting]: light type, direction, quality.
[Style / medium]: photoreal, cinematic, vector, 3D, illustration.
[Exact text]: "quoted copy here" in a named font/weight, placed [where].
Constraints: aspect ratio, resolution, must-include and must-avoid.

Here is the skeleton filled for a real job:

A launch banner for a fintech app aimed at first-time savers. Centered hero composition with generous negative space on the left for copy. A sleek smartphone floats at a slight angle showing a clean savings dashboard. Modern rooftop cafe at dusk, blurred city lights behind. Soft golden-hour backlight with a cool rim on the phone edge. Photorealistic, cinematic, editorial finish. Add a headline top-left reading "Save Smarter" in a bold condensed sans-serif, and a smaller line below reading "Round-ups that add up." Aspect ratio 16:9, 2K.

Why it works: every layer is explicit, the exact text is quoted and placed, and the constraints line locks ratio and resolution so Qwen plans the canvas correctly.

Aspect ratios & resolution

Qwen supports seven ratios and renders natively up to 2048x2048 (2K). Set the ratio with a UI preset or just type it into your constraints line.

RatioShapeTypical use
1:1SquareSocial posts, avatars, product tiles (1024x1024)
3:2LandscapePhotography, hero banners, editorial spreads
2:3PortraitPosters, book covers, magazine pages
4:3LandscapeSlides, infographics, classic photo frame
3:4PortraitMobile screens, portrait product shots
16:9Wide landscapeThumbnails, web headers, video stills (1536x864)
9:16Tall portraitStories, Reels, TikTok, phone wallpapers

Best for: request 2K whenever the image carries small text, charts, or fine detail — it keeps captions legible down to roughly 10px.

Text & rendering rules

Text is Qwen's flagship strength: legible small type across 12 languages and 20+ fonts, including formulas and subscripts. These rules are what actually hold up in practice.

RuleHow to write it
Quote exact wordsWrap copy in "double quotes" so Qwen renders it verbatim.
Name the font / weightSay "bold condensed sans-serif" or "elegant serif" — not just "nice font."
Put text earlyState the text instruction near the top so layout is planned around it.
Place itSay where: "top-left headline," "footer strip," "on the label."
MultilingualMix languages freely — strong bilingual Chinese + English layouts.
Formulas & subscriptsAsk for Greek letters, superscripts, and math notation directly.
Keep headlines tightQwen holds long copy well, but short headlines render cleanest.
A minimalist typographic poster on a warm cream background. Large centered headline reading "Deep Work" in a bold geometric sans-serif, with a thin subtitle beneath reading "Focus is a skill, not a mood." A small footer line reads "Session 07 — 90 minutes" in a monospaced font. Balanced margins, subtle paper texture, high-contrast ink. Aspect ratio 2:3, 2K.

Why it works: three text blocks are quoted, each has a named font and a defined position, and the ratio is fixed for a print-style poster.

Lighting modifiers

Lighting sets mood and depth. Drop one of these phrases into the lighting slot of your brief.

ModifierEffect
Three-point softboxClean, even studio light — product and portrait default.
Golden-hour backlightWarm glow behind subject, soft flares, editorial warmth.
High-keyBright, low-shadow, airy — beauty and lifestyle.
Low-keyDark, moody, deep shadows — drama and mystery.
Rim lightBright edge outline that separates subject from background.
RembrandtDirectional side light with a small cheek triangle — classic portrait.
Soft overcastFlat diffused daylight, gentle shadows — natural, honest.
NeonColored practical lights, magenta/cyan glow — nightlife, cyber.

Camera & lens modifiers

Naming a lens and aperture tells Qwen how much of the scene is sharp and how the perspective bends.

ModifierEffect
85mm f/1.4Flattering portrait compression with creamy background blur.
35mmNatural, documentary field of view — lifestyle and street.
24mm wideExpansive scenes, interiors, dramatic foreground.
100mm macroExtreme close-up detail — food, jewelry, textures.
Shallow DOFThin focus plane, soft background — isolates the subject.
Deep DOFEverything sharp front to back — landscapes, product flat-lays.
BokehSoft rounded out-of-focus highlights behind the subject.
Low angleCamera looking up — makes the subject powerful and tall.
Advertisement

Style & medium modifiers

Pick the medium last — it reframes everything above. One clear style word beats stacking five.

ModifierLook
PhotorealisticCamera-real detail, believable materials and light.
CinematicFilm color grade, wide contrast, story-driven framing.
3D renderClean CGI with soft global illumination and depth.
Flat vectorCrisp shapes, solid fills, no gradients — icons and UI.
WatercolorSoft washes, bleeding edges, paper grain.
IsometricAngled 3D-on-a-grid — diagrams, game scenes, product cutaways.
AnimeClean linework, cel shading, expressive characters.
Line artSingle-weight outlines, no fill — logos, tattoos, coloring pages.

Advanced settings

Most jobs need nothing beyond the prompt, but three controls help when you want more grip. Guidance scale (true_cfg_scale) runs roughly 1–10 with a default near 4 — higher tracks your prompt more literally, lower gives Qwen more freedom. Inference steps trade speed for detail — more steps mean finer results and slower renders. For editing, Qwen-Image-Edit accepts up to 3 reference images; state what stays first, then the single change, e.g. "keep the composition and lighting, change only Y." See the prompt templates for reusable skeletons.

Copy-paste example prompts

Five briefs across common jobs — lift, tweak the copy, ship.

1. Product hero

A premium product hero for a matte-black wireless earbud case, floating and slightly rotated on a smooth graphite pedestal. Studio three-point softbox lighting with a cool rim light along the top edge, subtle reflection beneath. Deep, rich shadows on a seamless charcoal gradient background, generous negative space on the right. Shot on 100mm macro, shallow depth of field, photorealistic, commercial finish. Aspect ratio 3:2, 2K.

Best for: e-commerce listings and landing-page heroes that need a clean, high-end object shot.

2. Text poster

A bold event poster for a jazz night on a deep navy background with grainy texture. Large headline reading "MIDNIGHT SESSIONS" in a tall condensed serif, gold foil finish. Below it a line reading "Live at The Blue Room — Fri 9PM" in a clean sans-serif. A small saxophone silhouette lit by warm neon glows lower-right. Balanced margins, high contrast, print-ready. Aspect ratio 2:3, 2K.

Best for: gig, sale, and announcement posters where the type is the hero.

3. Infographic

A clean single-column infographic titled "How Compounding Works" in a bold sans-serif at the top. Three numbered steps stacked vertically, each with a simple flat-vector icon on the left and a short caption on the right: "1. Invest early", "2. Reinvest returns", "3. Let time multiply it." A small line chart curving upward anchors the bottom with the label "Year 1 to Year 30." Soft mint and slate palette, generous whitespace, flat vector style. Aspect ratio 3:4, 2K.

Best for: explainer graphics, carousels, and slide decks that pack labeled data.

4. Editorial portrait

An editorial portrait of a confident woman in her 40s, a ceramic artist, standing in a sunlit studio with shelves of unglazed pottery behind her. She holds a half-finished bowl and looks just off-camera. Warm soft overcast daylight from a large window, gentle Rembrandt shadow on the far cheek. Shot on 85mm f/1.4, shallow depth of field, natural bokeh, photorealistic, magazine finish. Aspect ratio 4:3, 2K.

Best for: profile features, about pages, and brand storytelling with a human at the center.

5. Flat illustration

A friendly flat-vector illustration of a person working at a standing desk with a laptop, a coffee mug, and a potted plant. Warm coral, teal, and cream palette, solid fills with no gradients, clean rounded shapes and thin consistent outlines. Subtle geometric background pattern, plenty of negative space. Modern editorial illustration style for a blog header. Aspect ratio 16:9, 2K.

Best for: blog headers, app onboarding screens, and marketing spot illustrations.

Frequently Asked Questions

Does Qwen Image use --parameters?

No. Qwen Image 3.0 is not Midjourney and takes no --flags. You brief it in plain language like a creative director, and set the aspect ratio via a UI preset or by writing it into the prompt.

How do I set the aspect ratio?

Pick a preset in the UI (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3) or state it in plain text at the end of your prompt, for example "Aspect ratio 3:2, 2K." Qwen plans the layout around it.

What resolution can it output?

Qwen Image 3.0 Pro renders natively up to 2048x2048 (2K). Common presets include 1:1 at 1024x1024 and 16:9 at 1536x864. Ask for 2K when you need crisp small text or fine detail.

How do I get text to render correctly?

Wrap the exact words in double quotes, name the font and weight (for example "bold condensed sans-serif"), state where the text sits, and put the text instruction early so the layout is planned around it. Keep headlines tight.

What is true_cfg_scale?

true_cfg_scale is the guidance scale — how strictly Qwen follows your prompt. The usable range is about 1 to 10 with a default near 4. Higher values track the prompt more literally; lower values give the model more freedom.

How do I edit an existing image?

Use Qwen-Image-Edit with up to 3 reference images. State what stays the same first, then the single change, for example "Keep the composition, pose, and lighting. Change only the jacket color to forest green."

Where can I use Qwen Image 3.0?

Run it in Qwen Chat at chat.qwen.ai, via Alibaba Cloud Model Studio / DashScope API, or through third-party hosts like fal.ai, Replicate, Runware, and ComfyUI. The open weights (Apache-2.0) are on Hugging Face.

Advertisement