Home / Tools / Draw a UI

Draw a UI

Turns hand-drawn wireframes into working HTML with GPT-4 Vision.

Draw a UI preview
github.com Open Source

Upvotes

0

Pricing

Open Source

Category

Website Builders

In the database since

Oct 2026

Overview

About Draw a UI

Draw a UI is an open-source experiment by developer Sawyer Hood. It turns hand-drawn wireframes into HTML. You sketch a mockup on a canvas: boxes for layout, lines for navigation, scribbled labels. The app sends the sketch to OpenAI's GPT-4 Vision model. The model returns an HTML file styled with Tailwind CSS. It is a demo, not a product, and there is no hosted version. You run it on your computer with your OpenAI API key. Check the GitHub repo for setup steps, since details may change.

Key Features

1. Hand-drawn wireframe input

Sketch your interface with shapes, text, and freehand strokes on a full drawing canvas instead of writing prompts or dragging components.

2. Vision-model interpretation

Your drawing becomes an image read by GPT-4 Vision, so the model grasps layout, hierarchy, and labels visually instead of from text.

3. Single-file HTML output

Every generation returns one self-contained HTML file, making the result easy to preview, share, or drop into an existing project.

4. Tailwind CSS styling

The generated markup uses Tailwind utility classes, so the code stays readable and any developer can restyle it without untangling custom CSS.

5. tldraw drawing canvas

The sketch interface uses the open-source tldraw SDK, giving you polished tools — shapes, arrows, text, and freehand — out of the box.

6. Local, self-hosted setup

You clone the repo and add your OpenAI API key. Then run it on localhost:3000. Your sketches and credentials never pass through a third-party server.

7. MIT open-source license

The whole codebase is MIT licensed, so you can use, modify, and build on it — including commercially — with no licensing fees.

FAQ

What is Draw a UI used for?

It turns rough interface sketches into clickable HTML prototypes. Handy for validating a layout, briefing a developer, or exploring variations without hand-writing code.

How does it convert a drawing into code?

The app converts your canvas from SVG to PNG. It sends the image to GPT-4 Vision. The model returns one Tailwind-styled HTML file.

Do I need an OpenAI API key to use it?

Yes. There is no hosted version, so you supply your own key with GPT-4 Vision access, and OpenAI bills you per generation.

Do I need to know how to code?

Drawing needs no coding. Running it locally assumes basic terminal comfort. You will clone a repo and set an environment variable. Then run npm with Node 18.17 or newer.

Is the generated code production-ready?

No. The README calls this a demo — fine for mockups and prototypes, but the output needs review and cleanup before it could ship.

What is Terragon Labs?

It is the project the author names as this idea's "spiritual successor" — a more complete product built on the same drawing-to-interface concept.

Conclusion

Draw a UI is a small, honest open-source demo built on one simple idea. A rough sketch should be enough to start building a page. It will not replace a designer or front-end developer, and it was never meant to. But as a fast bridge from hand-drawn wireframe to clickable HTML, it earns a place in any builder's toolbox.

Newsletter

Tools like this change fast

New Website Builders tools launch every week — the newsletter keeps you ahead of what's worth trying.

Keep exploring

Similar AI tools

Ranked by the job they do, their capabilities, and how closely they fit this workflow.