Position is held in superposition while you move. Hold still and the system takes a measurement: the distribution collapses to one basis state, sampled under the Born rule, and your face resolves sharp again.
Video never leaves the device. Detection and rendering are local; nothing is recorded or uploaded.