AutoMovie API
    Preparing search index...

    Compile a surface network at one tier into its rest cage and a per-pose update of the cage's control positions.

    Each patch becomes a grid of its tier's rows and columns, sampled evenly in its own coordinates. A sweep patch evaluates its sweep, its radius under direction correspondence or its Cartesian section point under parameter correspondence averaged over each vertex's cell, which reaches half the spacing to the neighbouring sample on each side along u and v, as IAutoMovieSweepRig.sample states, so detail finer than the tier's grid fades instead of aliasing into ripples. A point sampled by several patches or sides takes, on each side, the finest spacing any of them gives there, and a side none of them reaches mirrors the other, so a point's position does not depend on patch order and a mirrored layout samples to a mirrored cage; a node on a sweep that no grid samples is the section point itself. A Coons patch takes its four sides' points and fills the interior as rigAutoMovieSurfaceCoons states; a bridge fills its rows with cubic Hermite curves between its two loops, as rigAutoMovieSurfaceBridge states; a transition band lays its quads out by layoutAutoMovieSurfaceTransition and samples every corner on its sweep. A network curve is compiled by rigAutoMovieSurfaceCurve, which regenerates its whole spline from sweep nodes and tangent planes evaluated with the same registry-resolved cell as their geometric targets. Observation-only sweep nodes keep a zero cell. This is a fixed local footprint derivative, not a differential of footprint size across the network. Vertices are shared by identity, never by distance: a point of a sweep is named by the sweep and its domain coordinates rounded to nine decimals, a curve point by the curve and its share, a node by its name.

    The patch grids alone decide the topology, so one network gives every input and every pose the same cage at a tier. A ring may only name points that patches made, and a bridge loop only sweep points that patches made, which catches sides that do not meet; an edge used twice in the same direction is refused as overlapping or misoriented patches.

    With fit (the default) the samples of sweeps and curves are targets the cage's Catmull-Clark limit passes through, solved by solveAutoMovieSubdivisionLimit; the interior vertices of Coons patches and bridges are blend controls of their own construction and keep their samples as control positions, because interpolating a blend's samples makes the limit surface ring between them where a loop turns sharply; pinned ring vertices keep their given positions. Without, controls are the samples themselves and the residual is 0. A ring that declares a perimeter is held to it at rest by holdAutoMovieSurfaceGirths, a radial factor along its sweep that every pose keeps. Every face corner carries (patch, s, t) and a UV value: the patch's rectangle in texture space, or the band t in [k / P, (k + 1) / P] of the unit square for patch k of P.

    Bindings make the rig posable exactly as poseAutoMovieSurfaceNetwork poses the network: bound sweeps by rigAutoMovieSweep, free nodes, curve through points, normals and directions and ring positions rigidly with their bones. Everything else is computed here once; pose refreshes only the bones, the bound sweeps, the curves, the samples, the blends, the pins and the fit, into buffers sized here.

    type IProps = {
        bindings?: IAutoMovieSurfaceNetworkBindings;
        construction?: ISamples;
        fit?: boolean;
        network: IAutoMovieSurfaceNetwork;
        preserveVolume?: boolean;
        rest?: readonly IAutoMovieResolvedBone[];
        tier: AutoMovieSurfaceTier;
    }
    Index

    Optional bone assignments by network names. Absence means all geometry stays at rest unless otherwise evaluated by the rig. Unknown names throw; any nonempty binding table requires rest bones.

    construction?: ISamples

    Optional rest inspection sample counts. Requested sweep observations and regenerated network curves are read after the girth solve; omission allocates no construction snapshot. Inspection is separate from the cage target fit.

    fit?: boolean

    Whether to solve control positions so the Catmull-Clark limit reaches authored surface targets. Omitted means true. False keeps samples as controls and reports zero fit residual; blend controls and pinned vertices remain held in either mode.

    Rest surface network, read without mutation. Compiled sweep rigs retain their source sweeps, so preserve defining data throughout the resulting rig's lifetime.

    preserveVolume?: boolean

    Whether bound sweep joint stations compensate scale to retain rest volume. Omitted means false; forwarded to sweep rig preparation.

    rest?: readonly IAutoMovieResolvedBone[]

    Resolved rest bones establishing binding transforms. Required when any binding table is nonempty; omitted with no bindings means no bone buffer is constructed.

    Sampling tier selecting each patch's authored row/column counts. Counts must be safe positive integers; invalid or unavailable counts throw during compilation.