Qwen Image 3 AI Image Generator

Made with Qwen Image 3

A handful of samples produced inside Fliki. Portrait scene, product still, motion frame, lifestyle portrait, painterly landscape, creator thumbnail, brand tile.

Prompt

A 30-year-old American woman with long auburn hair and a confident neutral expression stands in a sunlit downtown Chicago alleyway, wearing a bold red leather jacket and dark indigo jeans, warm late-afternoon light bouncing off red brick walls behind her, candid editorial fashion photography, photoreal, shot on 50mm at f/2, three-quarter framing, vertical 9:16 short-form fashion frame, no text, no labels.

Prompt

A premium frosted perfume bottle with a black cap stands center frame on a sunlit cream marble shelf next to a softly lit beeswax candle and a fresh eucalyptus sprig, warm window light from the right casting a soft realistic shadow under the bottle, captured on 85mm at f/4, ultra-photoreal luxury still life, vertical 9:16 e-commerce frame, no labels, no text, no writing.

Prompt

A 32-year-old American man with short brown hair in a charcoal training shirt and black shorts mid-stride on a tree-lined trail in a Boston park at sunrise, dynamic running motion, photoreal sports photography, golden backlight through the maples, shot on 70mm at f/2.8, vertical 9:16 motivational reel frame, no text.

Prompt

A 26-year-old American woman with shoulder-length brown hair and warm hazel eyes wears a beige cashmere cardigan and smiles softly at the camera while sitting at a sunlit small-town coffee shop window seat with exposed brick walls behind her, a latte in front of her, candid lifestyle photography, soft natural window light from the left, photoreal, vertical 9:16, no text, no captions, no labels.

Prompt

A solo American hiker in a bright orange shell jacket stands on a wooden footbridge in a misty Yosemite valley with granite cliffs rising on both sides and a teal river curving below, rendered as a painterly oil illustration with thick visible brushstrokes, warm amber and cool teal palette, vertical 9:16 painterly cinematic mood, no text.

Prompt

A confident 29-year-old American male tech creator with short curly dark-brown hair leans against a sleek matte-black desk in a modern studio, a glowing magenta neon line wrapping the back wall behind him, two large vertical monitors faintly visible on the right showing abstract green chart patterns, dramatic chiaroscuro lighting, photoreal subject, captured on 35mm at f/2.8, framed on the left third with clean negative space on the right, 16:9 ultra-photoreal cinematic editorial, no text, no overlay, no captions.

Prompt

A single premium ceramic candle in a textured cream jar centered on a soft sage green linen cloth with a sprig of fresh rosemary beside it, soft north-window light from the upper left casting a gentle long shadow, captured on 85mm at f/4, ultra-clean square 1:1 editorial brand tile, top-down view, no labels, no text, no writing.

Prompt

A 30-year-old American woman with long auburn hair and a confident neutral expression stands in a sunlit downtown Chicago alleyway, wearing a bold red leather jacket and dark indigo jeans, warm late-afternoon light bouncing off red brick walls behind her, candid editorial fashion photography, photoreal, shot on 50mm at f/2, three-quarter framing, vertical 9:16 short-form fashion frame, no text, no labels.

Prompt

A premium frosted perfume bottle with a black cap stands center frame on a sunlit cream marble shelf next to a softly lit beeswax candle and a fresh eucalyptus sprig, warm window light from the right casting a soft realistic shadow under the bottle, captured on 85mm at f/4, ultra-photoreal luxury still life, vertical 9:16 e-commerce frame, no labels, no text, no writing.

Prompt

A confident 29-year-old American male tech creator with short curly dark-brown hair leans against a sleek matte-black desk in a modern studio, a glowing magenta neon line wrapping the back wall behind him, two large vertical monitors faintly visible on the right showing abstract green chart patterns, dramatic chiaroscuro lighting, photoreal subject, captured on 35mm at f/2.8, framed on the left third with clean negative space on the right, 16:9 ultra-photoreal cinematic editorial, no text, no overlay, no captions.

Prompt

A 32-year-old American man with short brown hair in a charcoal training shirt and black shorts mid-stride on a tree-lined trail in a Boston park at sunrise, dynamic running motion, photoreal sports photography, golden backlight through the maples, shot on 70mm at f/2.8, vertical 9:16 motivational reel frame, no text.

Prompt

A single premium ceramic candle in a textured cream jar centered on a soft sage green linen cloth with a sprig of fresh rosemary beside it, soft north-window light from the upper left casting a gentle long shadow, captured on 85mm at f/4, ultra-clean square 1:1 editorial brand tile, top-down view, no labels, no text, no writing.

Prompt

A 26-year-old American woman with shoulder-length brown hair and warm hazel eyes wears a beige cashmere cardigan and smiles softly at the camera while sitting at a sunlit small-town coffee shop window seat with exposed brick walls behind her, a latte in front of her, candid lifestyle photography, soft natural window light from the left, photoreal, vertical 9:16, no text, no captions, no labels.

Prompt

A solo American hiker in a bright orange shell jacket stands on a wooden footbridge in a misty Yosemite valley with granite cliffs rising on both sides and a teal river curving below, rendered as a painterly oil illustration with thick visible brushstrokes, warm amber and cool teal palette, vertical 9:16 painterly cinematic mood, no text.

100M+

VIDEOS CREATED

12M+

USERS WORLDWIDE

80+

LANGUAGES SUPPORTED

What Qwen Image 3 delivers

Prompts up to 4.5K tokens

Most image models start dropping instructions after a paragraph. Qwen Image 3 accepts roughly 4.5K tokens of input, so a full spec - every headline, label, caption, and placement rule - lands in one generation instead of six rounds of patching.

Text legible down to 10 pixels

Dense newspaper pages, infographic grids, dashboard mockups, and small-print footers come out readable rather than as texture that looks like letters. This is what separates Qwen Image 3 from models that only render a headline well.

12 languages and 20+ fonts natively

Multilingual campaigns generate from one model instead of one model per script. Type sits correctly in each language, and named font styles give art directors a lever they normally only get inside a design tool.

Generation and editing in one model

Text-to-image, image-to-image, and prompt-driven editing all run off the same model. You do not hand off to a separate editor when the brief changes - iterate on the same image with the same prompt vocabulary.

Up to 3 reference images

Attach one to three inputs and a single instruction to compose them: put a person in a different scene, place a product into a styled set, or carry a character across shots. Faces, product details, and brand marks hold through the edit.

Structured content, not just pictures

Formulas, geometric diagrams, logical derivations, and multi-layer UI screens are explicit targets for this generation of the model. Useful for explainer visuals, course material, and product mockups where the content has to be correct as well as attractive.

Identity-preserving edits

Change a background, swap an outfit, or restyle a portrait without losing the face. The same holds for products and packaging, which is what makes it usable for catalog and campaign work rather than one-off art.

Straight onto the video timeline

Every image generated in Fliki lands in the same project as your voiceover, captions, and music. Make a poster or a scene still and drop it into the edit without exporting, re-uploading, or leaving the tab.

Hosted, no infrastructure

Qwen Image 3 is hosted-only - Alibaba has not published weights for this release. Fliki gives you access at 0.5 credits per image with no API keys, no GPU, and no queue of your own to manage.

How it works

How to generate an image with Qwen Image 3

Qwen Image 3 runs inside Fliki's AI image generator. Here's the six-step flow.

Fliki prompt input showing a long layout-heavy description for the Qwen Image 3 AI image generator

Step 1

Write your prompt

Qwen Image 3 takes prompts up to 4.5K tokens, so spell out the whole layout: headline, subhead, body copy, captions, and where each block sits. Put exact text in quotes. For an edit, describe the change instead: "swap the jacket to dark teal and leave everything else".

Fliki model selector dropdown with Qwen Image 3 chosen for AI image generation

Step 2

Select Qwen Image 3 as your model

Open Fliki's model selector and choose Qwen Image 3. Your request routes to Alibaba's third-generation image model at 0.5 credits per image.

Choose 16:9, 1:1, or 9:16 aspect ratio for Qwen Image 3 AI image generation on Fliki

Step 3

Pick your aspect ratio

Choose 1:1 for square, 16:9 for landscape, or 9:16 for vertical. Each ratio is composed natively rather than cropped, which matters when a layout has to stay balanced.

Attach up to three optional reference images to lock subject, product, and style with Qwen Image 3 on Fliki

Step 4

Add reference images (optional)

Skip this for pure text-to-image. For edits, attach up to 3 reference images to lock a character, product, or style - or to compose several inputs into one coherent shot.

Select output resolution for Qwen Image 3 AI image generation on Fliki

Step 5

Set the resolution

Higher resolutions buy you legibility. Qwen Image 3 holds text down to roughly 10 pixels, so dense infographics and small-print layouts survive at native output size.

Step 6

Generate

Hit Generate. Qwen Image 3 is paid-only on Fliki at 0.5 credits per image - upgrade your plan to access it alongside the rest of the premium catalog.

Qwen Image 3 FAQ

Frequently asked questions

Everything you need to know about generating and editing images with Qwen Image 3 inside Fliki.

Qwen Image 3 (Qwen-Image-3.0) is the third generation of Alibaba's Qwen image model, released in 2026. It handles text-to-image, image-to-image, and prompt-driven editing from one model, accepts prompts up to about 4.5K tokens, and natively supports 12 languages and more than 20 fonts.

Layout-heavy and typography-heavy work. Posters, infographics, dense text blocks, multilingual campaign assets, UI mockups, and diagrams - anything where the words and their placement have to be right, not just the picture around them.

No. Only Z Image Turbo and Flux 2 Klein are on Fliki's free tier. Qwen Image 3 requires a paid subscription and costs 0.5 credits per image.

Yes. Attach between one and three reference images and describe the change in plain language. The same model handles generation and editing, so you can go from a fresh generation to an edit without switching models.

Up to three. Reference images are optional - leave them off for pure text-to-image, or use them for edits, character consistency, product placement, and multi-image composition.

Qwen Image (the 2025 release) is a 20B-parameter open-weight MMDiT text-to-image model under Apache 2.0. Qwen Image 3 is the newer hosted generation: much longer prompt handling, broader language and font coverage, finer text rendering, and built-in editing. Alibaba has not published weights, a parameter count, or benchmark scores for this release.

No. Unlike earlier Qwen image models, Qwen Image 3 shipped without downloadable weights, a technical report, or a licence. It is available through hosted APIs only, which is how Fliki serves it.

GPT Image 2 leads on prompt reasoning and reference count (up to 16). Nano Banana 2 is the fast native-4K option. Qwen Image 3 is the one to pick when the image is mostly type and structure - long specs, small text, multilingual copy - at half the credit cost of GPT Image 2.

Still curious?

Try Fliki free in your browser, no credit card required.

Start free

Qwen Image 3 · Image generator

Generate your next image with Qwen Image 3.

Upgrade your Fliki plan to unlock Qwen Image 3 and the rest of the premium catalog.

Upgrade to generate

Free forever plan · No credit card required · Cancel anytime

Made with Qwen Image 3

What Qwen Image 3 delivers

Prompts up to 4.5K tokens

Text legible down to 10 pixels

12 languages and 20+ fonts natively

Generation and editing in one model

Up to 3 reference images

Structured content, not just pictures

Identity-preserving edits

Straight onto the video timeline

Hosted, no infrastructure

How to generate an image with Qwen Image 3

Write your prompt

Select Qwen Image 3 as your model

Pick your aspect ratio

Add reference images (optional)

Set the resolution

Generate

Built on the best AI models - ready inside Fliki

Frequently asked questions

What is Qwen Image 3?

What is Qwen Image 3 best at?

Is Qwen Image 3 free on Fliki?

Can Qwen Image 3 edit an existing image?

How many reference images does it accept?

How is Qwen Image 3 different from Qwen Image?

Is Qwen Image 3 open source?

How does Qwen Image 3 compare to GPT Image 2 and Nano Banana 2?

Discover more

Discover features

Generate your next image with Qwen Image 3.