⚡️ IMGPrompt Click in your language, copy out the English wording AI image and video models expect. ▶ Try it online · ⬇ Desktop app · Self-host guide · 简体中文 You know the look you are after — you just don't know that models call it Tyndall effect, film grain or Fisheye Lens. IMGPrompt is a prompt editor for AI art and video, built for people who don't think in English: it lays 5,000+ of those terms out by subject and…
You know the look you are after — you just don’t know that models call it Tyndall effect, film grain or Fisheye Lens. IMGPrompt is a prompt editor for AI art and video, built for people who don’t think in English: it lays 5,000+ of those terms out by subject and category, labelled in your native language and nearly all carrying a preview image. Click the ones you want and the English prompt for GPT-Image-2, Midjourney, FLUX or Veo assembles on the right.
Supported
What
Details
Target models
GPT-Image-2 · Nano Banana · Midjourney · FLUX · Seedance · Veo · Kling — also compatible with Stable Diffusion and DALL·E syntax
Languages
18, with the interface and the tag library both fully translated (direct links below). Arabic renders right-to-left
Run it in
Any modern browser, nothing to install · Windows (exe / msi) · macOS (Intel + Apple Silicon) · Linux (deb / AppImage / rpm)
Your data
There is no backend — your selection lives in the URL and in your own browser, and never reaches a server of ours
Built-in translation
Google · Microsoft Edge · Youdao · MyMemory, tried in that order until one answers
Pick a subject (①) → pick a category (②) → click the tags you want (③).
The English prompt assembles in the right-hand panel as you click. Hit Copy and paste it into your AI tool.
The filter you are viewing is written into the URL (?object=Character&attribute=Basic), so a bookmark or a shared link drops the next person on exactly the same view.
Features
5,040 bilingual tags across 16 subjects and 212 categories — lighting, composition, lenses, camera moves, art styles, materials, wardrobe — organised by hand, not scraped into one flat list.
A preview image on nearly every tag. Hover to see what a style actually looks like; click it for a lightbox with zoom, rotate and download.
Type in any language and see the English. The translate box works on free text, debounced 1.5s, and suggests curated tags that match what you typed.
Multicolor or monochrome tags — multicolor paints the grid in blocks of eight from a six-colour cycle, so your eye keeps its place in a long list. Plus light / dark / system theme; both choices persist and sync across tabs.
Open data. The tag library is plain JSON in src/app/data/prompt/prompt-<locale>.json — fork it, extend it, ship your own.
The translate box is the one thing that leaves your browser. What you type there is sent to free public endpoints, using no API key of yours. They rate-limit, and MyMemory refuses anything over 500 bytes — the app falls through the provider list, but a long paragraph can still come back untranslated.
Tag preview images are served from a CDN. The tag data itself ships with the app, but the hover previews need a network connection, including in the desktop build.
Self-host
The published image carries the finished 18-language site, so there is nothing to build:
docker run -d -p 5666:5666 --name imgprompt ghcr.io/rockbenben/img-prompt:latest
Open http://localhost:5666 and it is yours — on a NAS, a home server, or an air-gapped machine. A docker-compose.yml ships in the repo, and because the app is a plain static export, any static host works just as well. Node versions, build commands and Vercel steps are in the deployment guide.
Contributing
Tag contributions are the most useful kind, and they need no toolchain: every language is one JSON file under src/app/data/prompt/, so adding or correcting a term for your language is a single edit. For code, the stack is Next.js 16 + React 19 + TypeScript with Ant Design 6 and Tailwind CSS 4 — open an issue first if the change is a large one.
Star History
Credits and licensing
IMGPrompt’s own code is MIT. The prompt library 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.
评论