You found a picture with exactly the look you want. Image to prompt turns it into words you can reuse: the subject, style, light and camera, written so an image model can make something in the same style.
What a good image-to-prompt result describes
- Subject and what it is doing
- Setting
- Style: photo, film, 3D, illustration, painting
- Light and its direction
- Camera angle and lens feel
- Composition and main colours
- Mood and aspect ratio
Change only the first line to make something new in the same style. That is the main reason to do this: you keep the look and swap the content.
The fast way
Upload the picture to the free Image to Prompt tool and pick the model you'll use. It writes plain sentences for Gemini and ChatGPT, a single line with --ar for Midjourney, or keyword phrases for FLUX and Stable Diffusion.
Doing it by hand in ChatGPT or Gemini
Look at this image and write one prompt that would recreate it. Cover the subject and action, setting, style, lighting and its direction, camera angle and lens, composition, main colours, mood and aspect ratio. 60 to 130 words. Output only the prompt.
Tips
- Crop to the part you care about before uploading.
- If the result is generic, ask for "specific materials and colours".
- Treat the prompt as a starting point: change one detail at a time and compare.
- Use pictures you have the right to use as a reference, and don't try to copy a real person's likeness.
Looking for a starting point instead? The [image prompt library](/image-prompts) has prompts with the result each one makes.

