Invisible
OS

InvisibleOS began as a technical investigation: what does it feel like to interact with a user interface using nothing but your hands in space? No mouse, no keyboard, no touchscreen. Just the air between your body and a canvas, mediated by computer vision.

The physical setup strips away every familiar signal of computation. A bare white canvas, 12 by 24 inches. A projector offset to the side. A webcam mounted on top. No visible computer anywhere. The effect is a contract between two aesthetics that rarely meet: the futurism of gesture-controlled interfaces and the traditional blankness of an artist's canvas.

InvisibleOS parodies the logic of the modern tech product. It onboards you. It tutorials you. It makes you feel like you are learning something. And then, when you have finally navigated its puzzles and unlocked its sequences, it tells you there has been an error. All your data has been deleted. Would you like to speak to support? Yes or no — it does not matter. The system restarts. You calibrate again.

You approach the canvas. A skeleton of your hands appears on the surface.

The piece runs entirely in the browser via p5.js, with MediaPipe's hand tracking model processing webcam input in real time to extract 21 landmark points per hand. Those landmarks drive all UI interaction: pinch detection, depth estimation, drag and click events, and the calibration sequence.

The physical setup is deliberately minimal. A 12 by 24 inch canvas acts as the projection surface, facing the viewer directly. A short-throw projector offset to the side fills the canvas without the viewer casting shadows. A webcam mounted on top of the canvas looks outward, tracking the viewer's hands in the space between their body and the surface.

The audio layer uses synthesized tones and haptic-style feedback sounds timed to gesture events, reinforcing the sensation that your hands are genuinely touching something, even though they are touching nothing at all.