AutoMovie API
    Preparing search index...

    Namespace IAutoMovieMotionHumanPerformance

    A performance of people and props: their chained actions placed on the film clock, every cut checked against the actions' interrupt tables, and each person's displayed pose a pure function of film seconds.

    Construction resolves the composition with the engine's schedule resolver. Each intent's timing comes from its law's plan, made from the state the performer is in when the intent starts: the previous action's displayed pose and velocity at its cut or end, its posture and its planted feet. A cut (an intent's until) is legal only inside a window of the cut action's interrupt table for the next action; otherwise construction throws with the legal windows. A proximity cut is found by scanning the 24 Hz grid for the first frame within the distance and bisecting 20 times, about 1e-7 s, deterministic on a fixed grid whose last sample is the action's end; a distance never met leaves the action uncut. Given rules, the placed schedule is checked with the engine's legality check and a violation refuses the performance. An input equal to one the process resolved recently reads that resolution again (HumanPerformanceState.of): the resolution is a pure function of its input.

    Sampling a person at a second takes the active action's posture, resolves it to a pose, adds the decaying inertialization offset of the transition that entered it, re-solves every planted or gripping limb onto its goal so no contact slides, and clamps to the range of motion (HumanSegment). A person before their first action stands at their starting place. Additive layers (gaze, breathing) apply over every action without cutting it. While the schedule is resolved, a person's poses may read people whose actions are not placed yet, so nothing is cached until every action is placed; then caches and transition offsets start afresh.

    IConstructor
    IInput