Why Hand Gesture Control Webcam Demo Works Here
Full hand skeleton
Hand mode tracks knuckle and fingertip joints per hand, live, with adjustable detection confidence.
Simple pose gestures out of the box
Pose mode's Gesture Labels calls out held poses like 'Arms Up!' or 'T-Pose' — geometric checks, not a trained classifier, so they're easy to reason about.
Joint Angles for finer control
Pose mode can read elbow/knee bend angles in degrees, useful for threshold-based gesture logic.
Export tracking data
Motion Tracking Data Export can save tracked coordinates to CSV for hooking into your own gesture logic downstream.
Multi-hand and multi-person
Track more than one hand, or (with Multi-Person on) more than one person's pose, at once.
No install, runs locally
Everything runs in-browser — nothing uploaded, nothing to install, useful for quick interactive prototypes.
Frequently Asked Questions
Can this control my computer or other apps?
No, and it is worth being clear about that — it tracks and visualises hands and body poses in the browser, but it does not send input to your operating system or to other applications.
What does hand tracking actually detect?
Hand mode follows the joints of each hand, drawing a skeleton across the fingers and palm, and tracks both hands at once.
Does it recognise named gestures?
Pose mode labels a few positions geometrically — Arms Up and T-pose, for example — worked out from joint positions rather than a trained gesture classifier, so it recognises body posture rather than arbitrary hand signs.
Is there a delay between moving and seeing it?
It runs per frame on your own device, so latency is essentially your webcam plus one frame of processing. Lowering the Resolution setting speeds it up on older hardware.
Does it work in a phone browser?
Yes, using the front or rear camera, though hand tracking is the heaviest of the modes — drop the resolution if the frame rate suffers.
