OptionalaffordancesSemantic contact points the object declares: stackable tops, handles, sockets and hooks (IAutoMovieAffordance). Absent or null declares no contact points for affordance-based placement and support.
Id of an engine-resolved binary payload (the imported glTF/VRM file backing an imported model), or null when the geometry is fully described by parts (a generated model).
Authored mass, centre and contact properties, or null when absent. Null behavior is helper-specific: free-fall skips it, collision analysis uses documented defaults and centre estimation can still use primitive geometry. These available helpers currently have no product caller; this field alone enables no physical-feedback pipeline.
Stable model id cited by scene placements; poses and motions separately cite the skeleton id.
OptionalimagesImages the model carries for its own texture bindings, such as maps generated with its geometry. A binding's asset resolves against these ids before the project's assets, so these need no ledger record and no host loader. Omitted or empty, every binding names a project asset.
OptionalimportedCompiler-sealed execution facts for an imported appearance.
The viewer loads only this digest closure. Humanoid node indices come from ingest or authoritative VRM metadata, never a runtime name guess.
OptionallookHow the gaze channels of an expression turn this model's eyes.
Omitted, gaze drives morph targets and expression sinks, as "expression" does. "bone" turns the skeleton's leftEye and rightEye instead, by the ranges it declares.
Materials referenced by the parts.
Human / LLM readable name. Null if unnamed.
Whether the geometry was generated by automovie or imported by the user.
The geometry pieces and their material/rig bindings. At least one.
OptionalprofilesDeclarative profiles bound to this runtime model.
Omitted means the model proves no semantic engine capability. Imported appearance does not inherit traits unless its registered deterministic proxy supplies them.
The humanoid rig: bones, hierarchy, rest pose, and anatomical ROM. null for a non-articulated object/prop. A model with a skeleton is a character; one without is an object.
A 3D model: the riggable subject that poses, motions, and expressions drive.
This is automovie's central geometry container, and it deliberately unifies the two things the project animates:
skeleton: poses and motions articulate its bones;skeleton: null; it has geometry and materials but no joints, and is moved as a whole (placed in a scene, or parented to a character's bone as a held item).A model bundles its geometry (
parts), its rig (skeleton, optional), and itsmaterialsbehind one id; downstream layers cite it byidand never need to know whether it was generated from primitives or imported as a mesh (that isorigin). The LLM authors parts, materials, and the skeleton's ROM or a deterministic mesh generator; opaque vertex arrays are not authored by hand.