Three ways to turn a visual reference into a prompt

The image tool reads a single photograph, illustration or 3D render and describes the choices behind it: what the subject is and where it sits in frame, how the scene is lit, what the perspective and apparent focal length suggest, and how colour and texture behave across the picture. The output is written as a prompt you can paste into an image generator, plus a short negative prompt for the things you probably want to avoid.

It is most useful when you have a look you want to repeat with different content — a set of product shots that should feel like one campaign, a mood you saw in a film still, a lighting setup you want to study. Photographers use it to read a lighting and lens choice; illustrators use it to name a style they can only point at.

Images with one clear subject and a readable lighting direction work best. Busy collages, heavy watermarks, very low resolution files and screenshots of text tend to produce vaguer descriptions, because there is less consistent visual evidence to read. If your result feels generic, crop tighter on the part of the image you actually care about and run it again.

Read the image-to-prompt guide for what the analysis can and cannot infer.

Paste a public URL and Promptify fetches the page, reads its markup and stylesheets — including CSS custom properties — and renders the page so the section order comes from what is actually visible rather than from heading tags alone. From that evidence it writes a design brief: the palette with real hex values, the type scale and weights, spacing and radius conventions, component styling such as buttons and cards, and the order of the page's sections from header to footer.

You can ask for layout only, theme only, or both. Layout is the right choice when you are planning a page structure; theme is the right choice when you are setting up a design system and only need colour, type and component language.

Designers use it to describe a direction precisely in a brief instead of attaching a screenshot. Developers use it to bootstrap tokens before building. Product teams use it to state exactly what should stay and what should change in a rebuild. It describes a design direction — it is not a way to clone a competitor's site, and you should treat someone else's branding as theirs.

Pages that render everything client-side or block automated requests return a thinner reading than server-rendered pages. See how to create AI prompts for design inspiration.

Upload a poster, flyer, album cover, editorial spread or data graphic as an image or PDF. This tool reads composition the way a designer would: the underlying grid and margins, how the eye is meant to travel from the largest element downward, the relationship between headline, subhead and supporting text, the palette and how much of the surface each colour occupies, and the character of the type — weight, width, case, letter-spacing and the contrast between families.

It also describes graphic elements: rules and dividers, shapes and cut-outs, texture and grain, image treatment, and how much of the layout is left empty. For infographics it reads the chart types and how data is grouped and labelled.

Use it when you need to produce a family of layouts that share one visual system, when you are studying a poster style you admire, or when you want to hand a precise written brief to a collaborator or a generator. Flat, straight-on scans read better than photos of a poster on a wall, where perspective and uneven light distort the evidence.