Quick Answer
Upload an image (drag and drop, up to 3 MB), pick the output style — Midjourney prompt, Stable Diffusion tags with a negative line, or a plain description — and press Describe image. A vision model returns the prompt that recreates the picture: subject, style, lighting, composition, and camera language. 3 free runs a day on this site’s key; unlimited with your own Google Gemini key (the free AI Studio tier works) — the image goes only to the provider you chose, never to this site’s storage.
How It Works
Drop an image (PNG/JPG/WebP up to 3 MB) — it is read locally in your browser
Pick the output style: Midjourney prompt, Stable Diffusion tags with a negative line, or a plain description
Press Describe image: 3 free runs a day on this site’s Gemini key, or unlimited on your own free Gemini key
Copy the prompt or hand it to the Image Prompt Builder for refinement
Key Facts
- Midjourney, Stable Diffusion, or plain-description output styles
- Works on any image up to 3 MB — read locally, sent only to the chosen provider
- 3 free runs per day (resets UTC midnight); unlimited on your own Gemini key
- One-click handoff into the Image Prompt Builder to refine the result
Frequently Asked Questions
Where does my image go?
During the free trial it is sent to Google’s Gemini API to write the description and never stored here. With your own key it goes straight from your browser to Google — this site’s servers never receive it.
Why does it need a Gemini key specifically?
Vision (looking at an image) needs a multimodal model. In the BYOK path this tool uses Google’s Gemini, whose free AI Studio tier makes unlimited runs effectively free. OpenRouter keys work for text tools suite-wide, but the vision call here is Gemini-based.
How accurate is the recreation?
The prompt captures the describable elements — subject, style, lighting, composition. Exact reproductions of specific faces or logos are intentionally out of scope for image models, and prompt-level fidelity varies by model and seed.