PRIVATE · NO SIGN-UP · SINGLE IMAGE

Image to Prompt

Drop in an image. Get a clean, model-ready prompt and a negative prompt—without sending your image to a server.

Drop an image here or click to choose a file JPG, PNG, WebP or HEIC · up to 20 MB

Prompt format

Choose one image to begin.

Your prompt will appear here as soon as the image is analyzed.

Supported images

Upload one JPG, PNG, WebP or HEIC image up to 20 MB. Transparent images are supported. Animated GIF is not supported.

How it works

Choose an image and a target format. The page captions the image, reads its palette, contrast and composition, then writes a positive and negative prompt.

Model & privacy

Caption model: Xenova/vit-gpt2-image-captioning. It runs locally through Transformers.js. Model files are downloaded on first use; your image stays in this browser tab and is not uploaded by this page.

Free allowance

Unlimited local runs. There is no account, email gate or server quota. Your browser and device memory are the practical limits.

STRAIGHT ANSWERS

Image to prompt FAQ

Is this also a photo to prompt or picture to prompt tool?

Yes. Those names describe the same job: turning a photo, picture or other supported image into a reusable AI image prompt.

Which image generators can I use the prompt with?

Select Midjourney, Stable Diffusion, FLUX or DALL·E. The image prompt generator changes structure and parameters for the selected target.

Does my image leave the browser?

No. This page decodes and analyzes it locally. Your browser downloads the model code and model files, but the selected image is not sent to this site or the model host.

Why can the first result take longer?

The caption model must be downloaded and cached on the first run. Later runs can reuse the local browser cache.