CraftPrompts

Image to prompt generator

Drop in a photo, artwork or screenshot. A vision model describes it in the words image generators respond to, formatted for Midjourney, Stable Diffusion, Flux or a video model.

Write the prompt for
Prompt from image
Your prompt appears here. It usually takes 3 to 8 seconds.

Getting a closer match

  • Pick the right target. Midjourney and Stable Diffusion read comma-separated phrases; Flux, GPT Image and Nano Banana follow sentences. The same image gets a different prompt for each.
  • Use the focus box when you want the look rather than the content, e.g. “just the lighting and film grain”.
  • Pair words with references. In Midjourney add --sref with your image; see the parameters cheat sheet. In Stable Diffusion use img2img at a denoise of about 0.5 to 0.7.
  • Iterate one word at a time and keep the seed fixed so you can see what each change does.

Starting from an idea instead of an image? Use the AI image prompt generator or the Nano Banana prompt generator.

Questions

How does image to prompt work?

You upload a picture, and a vision-capable language model (served through Vercel AI Gateway) looks at it and describes what it sees in the vocabulary image generators respond to: subject, setting, medium, lighting, camera angle, composition, palette and mood. The description is then formatted for the generator you pick.

Will the prompt recreate the exact same image?

No. Generators start from random noise, so a prompt produces similar images, not copies. To get closer, add Midjourney’s --sref with the original as a style reference, use image-to-image in Stable Diffusion, or fix a seed while you tweak the words.

Is my photo stored?

No. The image is shrunk in your browser to at most 1280 pixels, sent once to generate the description, and not saved by CraftPrompts. Don’t upload images you don’t have the right to share with an AI service.

Can it describe people in photos?

It describes visible features like clothing, pose, expression and setting, but it will not identify who someone is. That keeps prompts usable and avoids recognising real people.

Is image to prompt free?

Yes, with a fair-use limit of 10 images per 10 minutes per visitor. No sign-up.

What is the difference between image to prompt and the image prompt generator?

Image to prompt starts from a picture you already have and reverse-engineers its prompt. The image prompt generator starts from an idea and helps you choose style, light and camera words. Many people use both: describe a reference image, then refine the result in the builder.

More tools in the studio