Image

Color and layout guesses. Not pixel-perfect HTML.

Drop a screenshot. We downsample, pick quantized HEX swatches, scan horizontal luminance bands, and emit a header / main / footer scaffold with CSS variables. No vision model. No OCR. No button detection. Treat the output as a sketch, then write the real UI.

No image selected

Drop screenshot here or browse — PNG, JPG, WEBP, SVG

Choose how aggressive the generated layout is — you can edit everything after copy.

Preview
Upload a screenshot to generate guidance.
<!-- Generate to see HTML -->
/* Generate to see CSS */

When to use this — and when not to

Use it to grab dimensions, a handful of quantized HEX swatches, a rough “how many horizontal bands” count, and a header/main/footer HTML+CSS scaffold with CSS variables.

Do not expect an AI coder, OCR, button detection, or pixel-perfect CSS. There is no model. Sampling is a downscaled canvas (long edge ~220px). Output is a sketch you rewrite against the real design system.

What the tool actually does

Draw the image to a canvas, quantize colors, split into horizontal luminance strips, pick a layout focus (mobile / neutral / desktop), emit analysis text plus semantic HTML and CSS. Images are not persisted. Only the layout-focus preference may sit in localStorage (droidxp-stc-target).

Worked example

A dashboard screenshot with a dark top bar and a light body often yields 2–3 suggested sections and a --stc-color-1 close to the bar. The generated <header> will not contain your actual nav labels — we never read the pixels as text.

Mistakes people make

  • Shipping the scaffold as the production layout.
  • Busy photos / gradients producing nonsense band counts. Ignore the count.
  • Huge images (over ~8192px) slowing the tab. Crop first.

Related

Image metadata viewer · App icon generator

Frequently asked questions

Is this AI?

No. Math on pixels. No server model.

Uploaded?

No. Canvas in this tab.

Will it read the labels in my screenshot?

No OCR. Type the copy yourself.

SVG?

The browser rasterizes it for sampling. Prefer PNG for huge files.