Skip to main content
  • Free
  • No signup
  • AI vision model
  • Image not stored

Image to Prompt Generator

Upload a photo or artwork and get a detailed prompt to recreate its look, formatted for Midjourney, FLUX, Nano Banana or Stable Diffusion, with a negative prompt where the model takes one.

Your image

Shrunk to 1,024 px in your browser, sent once to the AI model, never saved by IImagined.

What to capture

Capture

Prompt for

Add an image first. Up to 10 images a day.

Prompt for Midjourney

Your prompt and a breakdown of the image appear here.

Turn the look into a character

The AI Influencers course covers character creation, reference sets, LoRA training, content production and platform rules. Results depend on your workflow; no income or consistency outcome is promised. Start with the free Starter Kit: working files from the programs that open in your browser.

Get the free Starter KitOr see All Access: $99/mo

How it turns an image into a prompt

A free vision AI model looks at your image and writes a breakdown: subject, setting, composition, lighting, camera and lens, colour, style and mood, plus a written description and a list of things to avoid. The tool then formats that breakdown for the model you pick, using the same conventions as our AI Influencer Prompt Generator:

  • Midjourney: comma-separated description, then --ar set to the closest common ratio to your image's shape and --raw for a more literal, photographic result.
  • FLUX: natural sentences with the subject first. Black Forest Labs' prompting guide says FLUX.2 does not support negative prompts.
  • Nano Banana: full sentences plus the aspect ratio, as Google's image generation guide recommends describing the scene rather than listing keywords.
  • Stable Diffusion (SDXL and SD 3.5): comma-separated tags and a separate negative prompt.

Capture changes what the model looks at. "Style and lighting" leaves a [your subject] slot so you can put the look on something else. "The person's look" describes face, hair, skin and outfit in detail, a starting point for a consistent AI character.

What it does not do

  • It does not generate images. Paste the prompt into your own image model.
  • It does not recreate an image exactly. The same prompt gives different images on different models, and a new image every run.
  • It never names or identifies people, and it turns down nude or sexual images and any image of children or teenagers.
  • It does not store your image. IImagined keeps neither the image nor the prompt; the free model's provider may keep inputs under its own terms.

To design a character from scratch rather than from a photo, use the AI Influencer Prompt Generator. For more on reverse-engineering a style, read our image to prompt guide.

Questions

How does image to prompt work?

A vision AI model looks at your image and describes its subject, setting, composition, lighting, camera, colour, style and mood. The tool turns that description into a prompt in the format your image model expects: parameters for Midjourney, plain sentences for FLUX and Nano Banana, and comma-separated tags with a negative prompt for Stable Diffusion.

Will the prompt recreate my image exactly?

No. A prompt is a starting point. The same prompt gives different images on different models, and a different image each time on the same model. To get closer, edit the prompt, run it a few times, or add the original as a reference image where your model supports one.

Is my image stored?

IImagined does not save your image or the prompt. Your browser shrinks the image to at most 1,024 pixels and sends it once, through our server, to a free AI model on OpenRouter. Free model providers can keep what they are sent under their own terms, so do not upload private or sensitive photos.

Can it tell me who is in a photo?

No. The tool never names or identifies people, even well-known ones. It describes visible appearance only, such as hair, skin tone, clothing and expression. It also turns down nude or sexual images and any image of children or teenagers.

Which image models does it write prompts for?

Midjourney (with --ar set from your image's shape and --raw), FLUX, Nano Banana (Google's Gemini image models) and Stable Diffusion (SDXL and SD 3.5). Switch between them after the prompt is written: it reformats at once without reading the image again.

How many images can I convert?

Up to 10 a day, free and without an account. The tool runs on free AI models with a shared daily limit, so it can be busy at peak times; wait a minute and try again.

More free tools

All free tools