How It Works: Napkin Wireframing for the Digital Age
It’s a lightweight desktop app that lets you draw UI sketches by waving your finger in front of your webcam.
Hit Cmd+Shift+C
Start drawing in the air
Watch your gestures turn into simple line drawings overlaid on your screen
Raise your other hand to save
The sketch is automatically copied to clipboard.
Paste into Cursor (or anywhere else). Done!
No stylus required. No drawing tablet. No switching between applications.
Just your finger, your camera, and the ability to quickly communicate a bad sketch without opening a traditional design tool.
Notes from Building
Perfection is the enemy of getting your point across. It’s not a drawing app, it’s a communication app. The sketches are rough, disposable, and just detailed enough to convey spatial relationships. In testing, even crude diagrams helped LLMs understand layouts faster than detailed text descriptions. High fidelity would actually make this worse by encouraging you to perfect the drawing instead of quickly sharing the idea.
Good software should feel like telekinesis. Should be easy. Paint The Sky launches with a global hotkey, captures your drawing, and disappears back into the background. Everything goes straight to your system clipboard. Sketch something, paste it into Cursor, Slack, or wherever you need it. No file management, no upload steps, no switching between apps. Tools should enhance your workflow, not interrupt it.
Let’s try modes here too. The simple drawings works great if you’re already in the context of a layout. Cursor can usually figure out that context. But if you’re starting from scratch, a prompt can guide the LLM on what to do with your sketch. That’s why we added UI Mode, HTML Mode, and Drawing Mode. These modes wait for you to paste your sketch, then automatically add a prompt to your clipboard, telling Cursor whether you want to turn this into a UI, recreate the drawing exactly, or generate some simple HTML structure with it.
Who knew waving at your computer could actually work? We were skeptical about webcam-based hand detection, but MediaPipe works reliably even in low light conditions (sometimes testing needs to happen when the family is sleeping). No calibration, no Nintendo Power gloves, no special setup. This started as a solution for visual prompting, but reliable hand tracking via webcam could be applied to all sorts of interface problems. We’re just scratching the surface of gesture-based computing.