Promptary

Reverse-engineer any web image into a ready-to-use AI art prompt. Local-first, bring your own API key.

Promptary turns any image you find on the web into a prompt you can actually use. Most "image to prompt" tools give you a paragraph describing the picture. That is not a prompt — you still have to rewrite it. Promptary generates the prompt in the exact format your generator expects: • Stable Diffusion — comma-separated tags with weights, quality tags, and a negative prompt • FLUX — full sentences; adding "masterpiece" actually hurts here, so it doesn't • Midjourney — concise description with --ar / --style / --v parameters appended • GPT-Image — long, detailed paragraphs; it understands them well • nano banana — natural descriptive language • Seedream — Chinese, which is what it handles best Pick your target model once and every prompt comes out ready to paste. HOW IT WORKS Right-click any image on any site and choose "Reverse-engineer this image". A few seconds later you get a prompt you can copy, edit, or send straight into your generator. Hovering over an image also shows a small button, and Alt+Shift+P analyses whatever image your cursor is over. Prompts you like can be saved to your library, filed into categories, tagged, searched, and reused later. When you are ready to generate, one click inserts the prompt into your image tool's input box — no copy-pasting. BUILT FOR YOUR OWN LIBRARY • Categories (a tree) and tags (many-to-many) for organisation • Full-text search across titles, prompts, tags and notes • Trash with restore, so a mis-click is never fatal • Import and export everything as JSON, or export to a folder with images as separate files • Available in English, 简体中文, 繁體中文, 日本語 and 한국어 YOUR DATA STAYS YOURS No account. No sign-up. No analytics. No telemetry. Your images, prompts and tags are stored in your browser's local database and never uploaded to us. Your API key is stored separately from your library, and is deliberately excluded from exported backups. You choose how an image reaches a model. With your own API key, the request goes straight from your browser to the service you configured — it never touches us. If you use the built-in free channel instead, the request is relayed through a server we operate, which forwards it and does not retain your images. Open source — every claim above can be verified in the code: https://github.com/dingge001/promptary