retarget

Retarget: the carrying of one body's motion onto another body's own vocabulary of motion. A reading is taken first, and a reading is a small model-free vocabulary of what a camera saw — head yaw, pitch and roll, jaw open, each eye's openness, brow raise, mouth width, gaze across two axes, each hand's openness and a few named hand shapes, every figure a scalar from zero to one or a degree. The retarget maps that reading onto a rig's tokens: a yaw becomes a turn to the left or to the right, a jaw becomes one of the four mouth shapes, an eye becomes open or closed, a brow becomes a raise, a hand becomes one of its poses. Thresholds, dead zones and which part answers which scalar are laws as data, never numbers buried in code. Nothing of the camera itself enters the word. A reading in, tokens per part out.