An input with no buttons and no edges.
Every laptop has a camera and almost nothing treats it as a control surface. This was an experiment in what that interface has to do differently.
Touch and click are solved problems: the target is visible, the press is felt, the result is immediate. A gesture has none of that. There is no edge to aim at, no click to confirm, and no way to tell whether nothing happened because you moved wrongly or because the camera never saw you.
Render the machine seeing you
The hand skeleton is drawn live over the video — 21 tracked points, joints connected, updated every frame. Before any feature works, you can see that you are being tracked.
Separate what persists from what does not
Three stacked canvases: the skeleton is cleared every frame, drawn strokes persist, and effects live on their own layer. The video sits underneath, mirrored so movement matches intuition.
Give the same input several personalities
One tracker, five modes — playground, air draw, pop bubbles, emoji rain, and a repulsor. The mode bar is the only chrome on screen.
