Class GltfLoader

Fetches and parses a GLB file at url into GPU-ready resources.

Creates GPU vertex/index buffers and textures owned by the returned model; the caller must invoke model.destroy() to release them.

Parses an already-fetched GLB buffer into GPU-ready resources — the same work as GltfLoader.load without the fetch. Use this when the bytes arrive by another channel (e.g. an Electron IPC read) or when a Content Security Policy forbids fetching blob: URLs.

Loads a GLB as a flat list of static (non-skinned) Mesh primitives.

Skips skin/animation parsing entirely; joints/weights on the source vertices are ignored, so vertices are read in their authored (bind-pose) frame. Pair with a MeshRenderer for plain rendering, or pass keepData: true so the returned meshes can feed SdfVolume.fromMesh (which needs CPU-side vertex and index data).

Each primitive becomes one entry in meshes; the loader owns the GPU buffers and textures it allocates — call model.destroy() to release them.

Like GltfLoader.loadStatic but parses pre-fetched GLB bytes instead of fetching a URL — for environments whose CSP forbids fetch() of the model (e.g. the editor, which reads asset bytes over IPC).

Device-free decode of every static primitive in a GLB into packed StaticPrimitiveGeometry (48-byte vertex layout + indices), in glb mesh→primitive flat order. Does GLB parsing, Draco/meshopt/quantization decompression, the tangent solve (skipped when the material has no normal map), and the vertex pack — but touches no GPUDevice, so it can run in a Worker. The result feeds GltfLoader.loadStaticFromArrayBuffer via opts.packedGeometry.