This free image to prompt tool looks at a picture and writes the prompt that would recreate it. Upload a photo, an illustration or a screenshot and get a detailed prompt for ChatGPT, Gemini, Midjourney or Stable Diffusion, plus a shorter version and the details it picked up.
It’s the fastest way to learn how a style is described, to recreate an image you like in your own words, or to get a starting point you can edit.
Free image to prompt
Turn any image into a prompt
Upload a picture and get a detailed prompt that recreates it, ready to paste into ChatGPT, Gemini, Midjourney or Stable Diffusion. Your image is shrunk in your browser before it is sent, and nothing is stored.
Free, no sign-up. We keep nothing you paste in: it goes to the model and comes straight back to you.
How to use it
- Upload an image. It’s shrunk to 1,024 pixels in your browser before anything is sent, and we don’t store it.
- Pick the prompt style for the tool you’ll use.
- Optionally say what to change, like a different season, a watercolor look or a new background.
- Copy the prompt into your image tool, then tweak the words that matter most.
What makes a good image prompt
- Subject first. Say who or what the image is about before anything else.
- Setting and composition. Where it is, how close the camera is, and what’s in the background.
- Light and color. Golden hour, soft studio light, neon, muted pastels: these change the feel the most.
- Style or medium. Photo, oil painting, 3D render, flat illustration, pencil sketch.
- Mood and detail. Calm, dramatic, playful; highly detailed or minimal.
Natural language vs Midjourney vs Stable Diffusion
| Style | How it reads | Best for |
|---|---|---|
| Natural language | Full sentences describing the scene | ChatGPT, Gemini and Copilot image tools |
| Midjourney | A comma-separated description with parameters like –ar for the aspect ratio | Midjourney |
| Stable Diffusion | Keywords plus a separate negative prompt listing what to avoid | Stable Diffusion and similar open models |
What this tool won’t do
It describes what it sees in general terms. It won’t name real people, and it won’t tell an image model to copy a named artist, brand or character: it describes the style in plain words instead. Results are a starting point, so expect to adjust a few words to get closer to the original.
Want to build a prompt from scratch instead? Try our AI image prompt builder, or sharpen any prompt with the AI prompt improver.
Related on HOW TO // AI
- Free AI tools: more free generators, checkers and calculators.
- AI chatbot pricing compared: which plans include image generation, and the limits.
Get a practical AI workflow every Thursday: join the HOW TO // AI newsletter.
Frequently asked questions
How does image to prompt work?
The tool sends a shrunk copy of your image to an AI model that describes the subject, setting, lighting, colors and style, then writes them as a prompt in the format your image tool expects.
Is my image stored?
No. The image is resized in your browser, sent to the model to write the prompt, and not kept by us.
Will the prompt recreate the exact same image?
No image model reproduces a picture exactly from text. The prompt gets you close in subject, style and mood, and you can adjust the words to get closer.
Which prompt style should I pick?
Natural language for ChatGPT, Gemini and Copilot, the Midjourney style for Midjourney, and the Stable Diffusion style if your tool supports a negative prompt.