Image to prompt
Upload a picture and get the text prompt that would recreate it, written for the generator you use: Midjourney, ChatGPT, Nano Banana, Stable Diffusion or Flux.
Found an image whose look you want to reuse? Upload it here and get the prompt that describes it the way an image generator needs: the subject, the setting, the composition, the light, the colors and the style. Pick the generator you use, and the prompt comes back in that generator's format.
How it works
Your browser shrinks the picture to 768 pixels on the long side and sends that copy to our server. A vision model reads it and writes a prompt in the format you picked. It describes what is visible, in the order a generator weighs it: subject first, then setting, framing, lighting, palette, mood and medium. Within a few seconds the prompt appears under the form, ready to copy.
The format changes with the target, because generators read prompts differently:
| Target | What you get |
|---|---|
| Midjourney | One line of short visual phrases, most important first, then --ar with the picture's aspect ratio (and --style raw for photos) |
| ChatGPT | Two to four plain sentences starting with "Create" |
| Gemini / Nano Banana | Two to four plain sentences starting with "Generate", with the aspect ratio in words |
| Stable Diffusion | Comma-separated tags in order of importance, then a negative prompt |
| Flux | One natural-language paragraph that puts the subject first |
| Any generator | One detailed paragraph you can trim for whatever you use |
Getting a closer match
- Run the prompt, then compare. The first image will be close, not identical. Look at what differs (the light, the lens, the color of one object) and add that detail to the prompt in plain words.
- Keep the style words, change the subject. The most useful thing an image to prompt tool gives you is the style: the medium, the lighting and the palette. Swap the subject for your own and the new image keeps the look.
- Crop before you upload. If you only care about one part of the picture, crop to it. The model describes everything it sees, and a busy background ends up in the prompt.
- Edit a photo instead of recreating it. If the goal is your own photo in a new style, upload the photo to ChatGPT or Gemini directly and use one of our ChatGPT photo prompts or Nano Banana prompts, which keep your face unchanged.
What it will not do
It does not name people, brands or places unless their name is written in the picture, so a photo of a celebrity comes back as "a man in his forties in a gray suit", not a name. Text in the image is quoted only when it is short, such as a headline. Sexually explicit images and graphic violence are refused.
FAQ
Is this image to prompt generator free?
Yes, with no sign-up. To keep it free for everyone there is a fair-use limit of a few images every ten minutes per visitor. If you hit it, wait a little and try again.
Do you keep my image?
No. The reduced copy of the image goes to our server and to OpenAI's model that reads it, and is not stored or published. We keep only the text of the prompt, filed under a one-way fingerprint of the image, so that the same image gets the same answer without another model call. Don't upload private documents or photos of other people without their permission. See the privacy policy.
Which image formats work?
JPG, PNG, WebP and GIF (the first frame). Any size works: the picture is reduced in your browser before upload.
Can I turn the prompt back into an image here?
Not on this site. Paste the prompt into the generator you picked. To write an image prompt from a description instead of a picture, use the image prompt generator or the Midjourney prompt generator. For an idea to start from, try the art prompt generator.