Private, on-device gesture research
Gesture Proof
When a hand moves, what can you trust?
Eight visual studies make hand tracking legible: signal, confidence, intent, then controlled consequence.
Open a study. Try its deterministic fixture or start your device camera.
Camera starts only after your click.
- 21 landmarks
- 0 uploads
- 1 visible decision at a time
08 studies / local only / no hidden capture
A visible method
Do not skip
the evidence.
Each study answers one practical question before the next one is allowed to become more expressive.
-
Phase 01 / observation
See the signal before naming it.
Start with the tracker’s 21 points, their confidence, and the gaps where evidence is missing.
Open Landmark Explorer -
Phase 02 / temporal evidence
Make intent earn its name.
Ratios, hysteresis, confirmation, and release turn a noisy pose into a state that can safely persist.
-
Phase 03 / gesture vocabulary
Give different gestures one temporal contract.
A shared state matrix makes activation, hold, uncertainty, and release inspectable instead of mysterious.
Open Gesture State Matrix -
Phase 04 / material response
Let movement change a field, not the rules.
Effects receive normalized signals. Tracking stays separate, visible, and reversible.
Start where it moves
Four studies.
One visible chain.
005 / Motion Field
Velocity becomes a visible force.
Palm center → speed → bounded particle field ↗006 / Object Manipulation Bench
Confirm the gesture. Then transform one cube.
Locked state → translate / rotate / scale ↗007 / Aperture Field
Two hands make an optical field.
Reliable contact → refraction / pixelate / blur ↗008 / Aperture Object Set
Containment becomes a shared command.
Aperture preview → neutral → set transform ↗Experiment index
Every question.
Every result.
Observe the signal
Name intent
Compose vocabulary
Make movement material
- 005 Motion Field Can palm velocity drive an effect through a clean boundary? Validation / active
- 006 Object Manipulation Bench Can locked gestures transform one wireframe cube? Validation / active
- 007 Aperture Field Can two hands earn a selective optical field? Validation / active
- 008 Aperture Object Set Can a complete hand field select a transformable cube set? Validation / active