AutoMovie API
    Preparing search index...

    Variable DEFAULT_HUMANOID_ROMConst

    DEFAULT_HUMANOID_ROM: Partial<
        Record<AutoMovieHumanoidBone, IAutoMovieJointConstraint>,
    > = ...

    Default anatomical range-of-motion table for the normalized humanoid, keyed by AutoMovieHumanoidBone.

    These are approximate clinical norms (goniometry), in degrees, under the sign convention documented on IAutoMovieJointConstraint: flexion (+) / extension (−) about the sagittal axis, abduction (+) / adduction (−) about the frontal axis, external (+) / internal (−) about the limb's long axis. Left and right limbs share these clinical ranges; HUMANOID_REST_FRAME maps each onto its side of the T-pose rig, so a value means the same motion on both sides.

    A bone absent from this table (or an axis set to null) is treated as unconstrained / immobile on that axis. A skeleton may override any of these per-bone via IAutoMovieBone.constraint (a contortionist, a stylized non-human rig); this table is only the fallback.

    These numbers are a deliberate, documented baseline: good enough to reject grossly impossible poses (a backward elbow, a hyper-extended knee). Sourcing exact per-population norms is tracked as future work.

    Ball-joint swingDeg caps use AAOS normal pure-plane maxima that must stay legal on their own: shoulder flexion/abduction 180 degrees, hip flexion 120 degrees. The cone is a conservative combined flexion+abduction cap, not a population-specific reach envelope.

    The HIP cone is live (its 120° cap sits below the flexion+abduction corner, so max-flexion + max-abduction trips it); the SHOULDER cone is deliberate headroom (#1058, decision 310): the swing metric never exceeds 180°, and a pure-plane 180° (arm straight overhead) must stay legal, so no cap below 180 is possible without rejecting canonical overhead poses. The value documents the sourced anatomical bound rather than gating anything.

    The shoulders are gated by swing and twist (IAutoMovieJointConstraint.swingTwist): the limb may point wherever its flexion and abduction intervals (and the cone) reach, and may turn about its own axis within an interval measured from the rig's rest pose. Their clinical twist interval is then a reading only. Read as three angles, the shoulder's twist range is fixed in a frame that turns once about the arm pointing straight ahead, so two arms a fraction of a degree apart there could lie one inside and one far outside it, and an arm on a table with its forearm up (flexion 135, abduction 71) read only 22 degrees of external rotation left. The shoulder's rest is the T-pose, 90 degrees of abduction, where its rotation is measured clinically (external 90, internal 70, as relayed for the AAOS norms and not checked against a primary source here). Measured from that rest, the turn a raised or a lowered arm needs is larger: a hand behind the head reads 128 degrees and a hand clasped behind the back -104 (the reaches of test_motion_fit_chain_reach, which the three-angle box admitted). So the interval is [-110, 135], provisional: one interval for every swing, where the anatomical limit depends on the swing (Codman's paradox: the turn a swing adds makes one interval of clinical rotation, internal 70 and external 90 at 90 degrees of abduction, refuse a hand behind the head by 38 degrees and one behind the back by 34). Their swing is capped at REACH from the T-pose, so the rest direction's antipode, the arm pointing through the chest, where the twist has no value, lies 50 degrees outside the range. The hips keep the three-angle gate: their twist is proximal, about the rest axis, and measured from rest by swing and twist the turns their box admits run from -106 to 91 degrees (deep flexion with abduction), which no one interval of the clinical 45 each way holds: a deadlift's planted feet slid 33.6 mm under it.