English | 中文 | 繁体中文 | Português | Italiano | Deutsch | Русский | Español | Français | 日本語 | 한국어 | हिन्दी | العربية | বাংলা | Bahasa Indonesia | Türkçe | Tiếng Việt | ภาษาไทย
Make AI Art Accessible in Your Native Language
Built with
IMGPrompt is an AI image/video prompt editor for non-native English speakers. It turns native-language clicks into clean English prompts ready for GPT-Image-2, Nano Banana, Midjourney, FLUX, Seedance, Veo, and Kling (also compatible with Stable Diffusion and DALL·E syntax).
The core power of IMGPrompt lies in its 5000+ curated prompt library. Forget wrestling with English vocabulary — browse and click keywords in your own language (Chinese, Hindi, Spanish, and 15 more), and the app assembles the precise English prompt AI models actually understand.
Experience IMGPrompt now - Make AI Art Accessible
- Language is no barrier. Pick tags in your native language, get English output.
- 5000+ tags, curated. Styles, artists, lighting, composition, camera angles — organized, not dumped.
- Visual discovery. Hover any tag to see a preview image; click the preview to zoom in a full lightbox (rotate / download included).
- Shareable selections. The filter you're viewing lives in the URL — send the link, land on the same view.
- Built for GPT-Image-2 and beyond. One workflow covers image and video models alike — GPT-Image-2, Nano Banana, Midjourney, FLUX, Seedance, Veo, Kling.
- Desktop or browser. Use the web app or download the native client for Windows, macOS (Intel + Apple Silicon), and Linux.
A three-step flow, numbered right in the UI so you always know where you are:
- Pick an Object (e.g., Character, Scene, Style) — section ①.
- Pick an Attribute under that object (e.g., Lighting, Pose, Camera) — section ②.
- Click the tags you want — section ③. Your selection builds the English prompt in the side panel. One-click copy, paste into your AI tool.
The URL updates as you browse (?object=Character&attribute=Lighting), so bookmarks and shared links take collaborators straight to the exact filter you're using.
Thousands of prompt keywords, organized by object and attribute. Verify meaning in your language, ship the precise English term models recognize. Data lives in open JSON at src/app/data/prompt/prompt-<locale>.json—fork it, extend it, it's yours.
Hover any tag → an example image shows up in the tooltip. Click the preview → a full lightbox opens with zoom, rotate, and download controls. Closing it keeps your browsing state intact. WYSIWYG prompt picking — especially useful when comparing GPT-Image-2 styles where the label alone tells you nothing.
Toggle between two visual modes from the tag area:
- Multicolor — 10-color antd palette cycled across tags, so large categories read as distinct groups at a glance.
- Monochrome — calm, uniform tag styling. Same data, less visual noise.
Your choice persists across sessions and syncs across browser tabs.
The right-side output panel is the working surface:
- Inline template buttons — one click to append "Portrait Lighting" or "High Quality" boilerplate. Copy a Negative-prompt template in one tap.
- Auto-translate while you type — debounced 1.5s. Type in any language, see the English equivalent surfaced below without clicking a button.
- Tag suggestions — as you type, exact-match and fuzzy-match tags appear so you can swap free text for a curated tag.
- Deduplication + random color — cleans up duplicated tokens, swaps a random color token for quick creative breaks.
- One-click copy and clear — side-by-side for speed.
Your currently active Object and Attribute filter lives in the URL. Bookmark the view you use daily. Send a link to a teammate and they land on the same filter. No accounts, no backend.
Auto-follows your OS theme (via next-themes), or override it manually. All tag colors, previews, and borders adapt.
Prefer a native app? Windows, macOS, and Linux builds ship on the Releases page.
Arabic, Bengali, Chinese (Simplified + Traditional), English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese, Russian, Spanish, Thai, Turkish, Vietnamese. Both the interface and the prompt database translate. RTL languages (Arabic) render right-to-left automatically.
A translation utility is also built in for free-form text, backed by Google / Youdao / MyMemory.
- Next.js 16 with the App Router and React Compiler
- React 19.2 + TypeScript 5
- Ant Design 6 (cssVar mode) + Tailwind CSS 4
- next-intl 4.9 for i18n and static generation per locale
- next-themes for dark mode
- Default locale:
zh. Root/redirects to/zh.
For detailed installation, local development, and deployment instructions (including Docker support), see the deployment guide.
Quick start:
npm install
npm run dev # http://localhost:3000
npm run build # production build
npm run build:lang # per-locale static build (scripts/buildWithLang.js)
npm run lintNode >=20.9.0 required.
Prompt data is aggregated and curated from several open sources — huge thanks to:
- EvoLinkAI/awesome-gpt-image-2-prompts — GPT-Image-2 prompt examples, licensed under CC BY 4.0. Entries are localized into 18 languages; per-entry author credits are preserved in the data file.
- Physton/sd-webui-prompt-all-in-one — baseline tag taxonomy (AGPL-3.0).
- promptoMANIA — structured keyword inspiration.
Pull requests are welcome. For major changes, open an issue first to discuss direction. Prompt data contributions (adding or refining tags for a language) are especially appreciated—the files are plain JSON, no tooling required.