Cycling through all six. Tap any point to stop.

Measured on six things

AI image-generation and editing studio for turning ideas into visuals. Create new images, refine references, explore transparent assets, and develop detailed prompts for portraits, products, landscapes, and visual storytelling.

Security

HTTPS not enforced, no HSTS, no CSP, no frame protection — 4/12 on the header checklist. No SOC 2, ISO, or security page. Standard for an early indie site, but thin.

Screenshots

1 / 1

Qwen Image 2.1

About this app

QwenImage AI is an independent web studio built around Qwen Image 2.1, a 7B model from the Qwen team that combines text-to-image generation and image editing in one workflow. The site stands out for its practical, no-nonsense approach: it offers a prompt playground, a curated gallery of visual studies, and a detailed guide covering prompt structure, editing with references, transparent RGBA assets, and multi-reference composition. You can generate images at 1K or 2K resolution using a credit-based system (28 credits at 1K, 53 at 2K), and failed generations are automatically refunded. The studio supports text-to-image prompts and drafting edit instructions, though live editing via the endpoint is still in preview. Pricing is straightforward with Basic, Pro, and Max tiers starting at $9.90/month. While the gallery images are original illustrations rather than model outputs, the site is honest about its limitations and focuses on helping you craft better prompts before you commit credits.

Key features

1Prompt playground for Qwen Image 2.1 with 1K/2K generation at 28/53 credits per image.
2Detailed prompt guide covering subject, scene, references, RGBA assets, and multi-reference composition.
3Transparent credit pricing with automatic refunds on failed generations.

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Graphic designers and digital artists

Pain points

Struggling with vague prompts that yield inconsistent AI images

Primary needs

A reliable studio to test prompts and generate high-res assets quickly

Demographic

Marketing and social media managers

Pain points

Need fast, on-brand visuals without hiring a designer

Primary needs

Transparent product cutouts and editable reference-based images

Demographic

AI enthusiasts and indie developers

Pain points

Fragmented tools for generation and editing; unclear pricing

Primary needs

An all-in-one playground with transparent credits and model documentation

Scout Review

Written by AI from measured evidence, scored out of 100.

QwenImage AI is a well-written, honestly-positioned front-end for Qwen Image 2.1 with a genuinely good prompt guide and transparent credit pricing. The design has editorial character, desktop performance is excellent, and accessibility is near-clean. But the studio is incomplete — editing isn't wired up, the gallery doesn't show model outputs, and security headers are largely absent. It's a promising shell around someone else's model, and the shell needs to become a product before the model provider eats its lunch.

Measure by measure

Usability

Prompt guide is genuinely useful and pricing is transparent, but editing isn't wired to the endpoint yet and the studio requires sign-in before you can try anything.

Design

Clean editorial layout with serif type and a muted gallery palette. Honest labelling of gallery as illustrations, not model outputs, but that also weakens the visual proof.

Speed

Desktop Lighthouse 98/100 and CLS 0 are strong; mobile 81/100 with a 4.9s LCP drags it down. Fast enough, but the mobile first paint is slow.

Security

HTTPS not enforced, no HSTS, no CSP, no frame protection — 4/12 on the header checklist. No SOC 2, ISO, or security page. Standard for an early indie site, but thin.

Accessibility

Lighthouse 96/100 with only a contrast failure. No WCAG statement or a11y page, but the site is responsive and readable. Solid for the stage.

Growth

Real problem (fragmented generation/editing, opaque credits) with a clear wedge in transparent pricing and prompt guidance, but the product is a thin wrapper on Qwen's model with no moat.

Improvement areas

Editing is not connected to the generation endpoint — users must copy instructions elsewhere.
Gallery shows original illustrations, not actual model outputs, so there's no visual proof.
Security headers are largely absent: no HSTS, CSP, frame protection, or DMARC.

Design & Usability

The site has a distinct editorial voice — serif headlines, numbered guide sections, a muted gallery palette — that reads more like a design magazine than a typical AI tool landing page. That's a genuine strength. The prompt guide is the best part of the product: it walks through subject, scene, and detail with concrete examples, and the pricing page is refreshingly honest about credit costs and auto-refunds. The problem is that the studio itself is half-built. Editing is explicitly 'not connected to the generation endpoint yet,' so users are told to copy their edit instructions into a different tool. The gallery is labelled as original illustrations rather than model outputs, which is honest but means you can't actually see what the model produces before you pay. Sign-in is required before any prompt submission, adding friction to what should be a try-before-you-buy flow.

Speed & Security

Measured performance is split: desktop Lighthouse hits 98/100 with zero layout shift, but mobile lands at 81/100 with a 4.9-second LCP — the first paint on a phone is slow enough to notice. CLS of 0 and TBT of 112ms suggest the page is well-built once it loads, so the mobile issue is likely asset weight rather than a structural problem. Security is the weakest measured dimension: the HTTP probe found HTTPS not enforced, no HSTS, no CSP, no frame protection, and no DMARC, landing at 4/12 on the header checklist. There's no SOC 2, ISO, or security page anywhere on the site. For an early indie studio handling payment and user accounts, that's a gap worth closing — at minimum, enforce HTTPS and add HSTS.

Accessibility & Growth

Accessibility scores well at 96/100 on Lighthouse, with a single failing audit for insufficient color contrast. There's no WCAG statement or dedicated accessibility page, but the site is responsive, keyboard-navigable, and readable — solid for an early-stage product. Growth is the harder call. The problem it names is real: AI image tools are fragmented between generation and editing, and credit pricing is often opaque. QwenImage AI's wedge is transparent credits, auto-refunds on failures, and a genuinely useful prompt guide. But the product is a thin wrapper around Qwen's 7B model, and the alternatives — Midjourney and Leonardo AI — are established with their own editing workflows. The domain is barely a month old, there are no traction signals, and the moment Qwen ships a first-party studio, the wrapper advantage evaporates. The idea can grow, but only if it builds something the model provider can't copy.

Conclusion

If you're a designer who wants a clear prompt guide and honest credit pricing before committing to a generation, this is a reasonable place to start — just don't expect the editing workflow to work yet. If you need a complete generation-and-edit loop today, Midjourney or Leonardo will get you further. The site's honesty about its limitations is refreshing, but honesty isn't a feature. Wire up the editing endpoint, show real model outputs in the gallery, and enforce HTTPS — then this becomes a product worth recommending.

Suggestions for improvement

Connect the editing endpoint so the studio actually does what the guide describes.
Replace the illustration gallery with real Qwen Image 2.1 outputs, clearly labelled.
Enforce HTTPS, add HSTS and a basic CSP — 4/12 on headers is too thin for a paid product.
Fix the contrast failure flagged by Lighthouse and add a short accessibility statement.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	Qwen Image 2.1	Midjourney	Leonardo AI
Transparent credit pricing with per-image cost	Yes — 28 credits at 1K, 53 at 2K, auto-refund on failure	Subscription tiers with GPU-hour based limits	Token-based credits with a free tier
Native transparent RGBA output	Claimed via Qwen Image 2.1 model; guide covers RGBA prompts	No native transparent output	Limited; not a core workflow
Multi-reference composition	Documented up to 10 reference images; drafting only in studio	Character and style references, limited multi-subject	Image guidance and reference features
Live image editing in-app	Not connected to endpoint yet — draft only	Vary, pan, zoom, and inpainting tools	Canvas editing and inpainting
Prompt guidance and documentation	Detailed three-move guide with practical notes	Community-driven docs and prompt resources	Tutorials and community prompt library