Your camera becomes the controller.
Move a 3D object with hand gestures, see your joint angles and balance as you move, and watch blinks and mood detected live. Computer vision running at camera speed, right in your browser.
Runs on your device · your video never leaves your browser and nothing is recorded
Turn on your camera to start
Tracking runs locally with on-device AI models (about 5 MB, downloaded once). Good light and a plain background work best.
Turn on the camera, then use the gestures below
- SteerOpen palmMove your open hand to rotate the object
- Grab & movePinchPinch thumb and index, then drag
- ScalePinch ×2Pinch with both hands and pull apart
- Next shapeVictoryShow two fingers
- Next colourThumbs upThumbs up
- Air drawPointPoint with your index finger to draw
- ResetFistHold a fist for one second
Finger positions
no handRaise a hand in front of the camera.
Readout
- Hands
- 0
- Fingers up
- —
- Pinch gap
- —
- Action
- —
- Shape
- Torus knot
- Colour
What happens on every frame
- 01
Capture
Webcam frames
- 02
Detect landmarks
21 points per hand
- 03
Interpret
Finger curl → gesture
- 04
Smooth
Debounce & smooth
- 05
Act
Drive the 3D object
Where businesses use this
Touchless kiosks
Browse menus and product catalogues in malls and clinics without touching a screen.
Showrooms & 3D configurators
Spin a car, a villa model or a product in 3D with a wave of the hand.
Sign & gesture input
Gesture commands for accessibility, AR and hands-busy work like surgery or assembly.
How it works: Google MediaPipe landmark models run in your browser (WebAssembly and WebGL) and return hand, body or face points for every video frame. Gestures, joint angles, stability, blinks and mood are computed from those points with rules we tune per project. This is a demo: mood is read from facial expression only, and the readings are indicative, not medical or psychological assessments. Nothing is uploaded or stored.