An open-source, production-oriented Qwen Image 3.0 prompt cookbook created and maintained by the flaq.ai team.
SEO topics: Qwen Image 3.0 prompts, Qwen Image prompt examples, AI image generation prompts, multilingual poster design, AI product photography, cinematic realism, e-commerce images, infographics, scientific diagrams, technical explainers, historical reconstruction, document publishing, character consistency, UI mockups, game assets, typography, image editing, and reusable prompt templates.
README languages: 简体中文 · English · 日本語 · 한국어 · Español · Français · Deutsch · Português · Italiano · Русский · العربية · ไทย · Bahasa Indonesia · Tiếng Việt · 繁體中文
→ Open the English prompt library: featured prompt, generated image, and all 144 recipes
Prompt index · Production templates · Use-case matrix · Advanced guide · Multilingual guide · Translation maintenance
- 144 original, copy-ready prompt recipes across 18 practical categories;
- 15 complete language editions: 18 canonical English files plus 252 locale-specific category files, for 270 browsable collection files in total;
- 16 reusable production templates and a 60+ deliverable use-case matrix;
- 19 newly generated example images, including 12 localized visual samples, with their production briefs;
- a structured prompt framework for long, text-heavy, and multi-panel compositions;
- practical guidance for multilingual typography, exact copy, character consistency, and controlled editing;
- contribution and quality-review rules designed to keep the collection useful and original.
Use the library for brand campaigns, e-commerce listings, restaurant menus, social posts, YouTube thumbnails, explainers, teaching materials, presentations, portraits, cinematic photography, storyboards, architecture, fashion lookbooks, destination marketing, editorial typography, white papers, manuals, heritage interpretation, scientific cutaways, game art, industrial concepts, and controlled image edits. Every recipe defines the deliverable, visual hierarchy, exact copy, constraints, and common failure modes—not just a loose list of keywords.
Qwen's July 2026 announcement describes Qwen Image 3.0 as supporting prompts up to roughly 4.5K tokens, text around 10px in showcased results, 12-language rendering, multiple fonts, 100+ visual styles, and interface-oriented compositions. These capabilities are not a guarantee of error-free output: always verify copy, numbers, diagrams, and high-stakes facts. See the official Qwen Image 3.0 announcement.
Each language README opens its own localized prompt page. The featured prompt and the displayed image are paired, so readers can copy the exact brief behind the visual and then continue into the complete 144-recipe library.
Localized editions are full recipe files, not redirect pages. Every one of the 144 recipes adds a target-language directive and script-specific typography QA covering CJK glyph variants, Hangul composition, Arabic RTL flow, Thai marks, Vietnamese diacritics, Cyrillic/Latin separation, accents, punctuation, and line breaking as appropriate.
| 한국어 | Español | Français |
|---|---|---|
 |
 |
 |
| Copy the Korean prompt | Copy the Spanish prompt | Copy the French prompt |
| العربية | Tiếng Việt | 繁體中文 |
|---|---|---|
 |
 |
 |
| Copy the Arabic prompt | Copy the Vietnamese prompt | Copy the Traditional Chinese prompt |
Browse all image-to-prompt pairs in the asset production notes.
Use case: [poster / product / infographic / character / UI / edit]
Asset type: [final destination and aspect ratio]
Goal: [the job this image must accomplish]
Audience: [primary viewer]
Scene/backdrop: [place, time, environment]
Subject: [subjects, actions, counts, relationships]
Style/medium: [photo / illustration / 3D / collage / UI]
Composition/framing: [camera, hierarchy, whitespace, reading order]
Lighting/mood: [light and emotion]
Color palette: [primary, accent, prohibited colors]
Materials/textures: [observable surface details]
Text (verbatim): [quoted lines only]
Data/labels: [numbers, units, labels, mappings]
Constraints: [must preserve, must match, must not change]
Avoid: [watermarks, extra copy, distortion, protected brands]
- Choose a category based on the final deliverable—not only the visual style.
- Copy one recipe and replace every
[bracketed variable]. - Put required on-image copy in
Text (verbatim)and keep each line inside quotes. - Move non-negotiable facts, counts, identity features, and layout rules into
Constraints. - Use
Avoidfor plausible failures such as extra copy, warped products, inconsistent characters, or watermarks. - Generate, review at full size, and revise one failure at a time. Human-check all spelling, numbers, diagrams, and sensitive claims.
For image editing, explicitly say: Change only [target]. Preserve [identity, composition, background, lighting, material, and text].
The project applies content-design principles developed by the flaq.ai team in Awesome SEO Writing Skill: search intent first, facts separated from creative judgment, audit before delivery, and complete local artifacts. Prompt text, fictional brands, visual briefs, and repository images were authored for this collection.
This is an independent community project and is not endorsed by Qwen or Alibaba Cloud. Qwen-related names belong to their respective owners.
flaq.ai provides a unified Agentic API for image, video, and language models. This repository is maintained by the flaq.ai team as an open-source resource for practical visual production.
Developers, agent builders, creators, reviewers, and educators can join the flaq.ai Affiliate Program, create a referral link, recommend flaq.ai services to relevant audiences, and earn commission from eligible referred orders.
- 20% on a referred user’s first valid paid order;
- 10% on following valid paid orders made within 60 days after that user registers;
- referral links and activity can be managed from the affiliate workspace.
Eligibility, attribution, refunds, chargebacks, risk reviews, payout readiness, and commission are governed by the active affiliate policy and agreement. Check the official page before publishing promotional content.
MIT © 2026 Flaq AI