WOM1/WOM2 had a single mesh with one texture, which lost the multi-submesh
structure of complex M2 models (body+hair+eyes+armor each need different
textures and blend modes).
WOM3 adds a Batch array: each batch has indexStart/indexCount + a textureIndex
into texturePaths + blendMode + flags. Loader is fully backward compatible:
WOM1/WOM2 files still load, and WOM3 with no batches block falls back to a
single full-mesh batch. fromM2 now extracts batches with materials, and toM2
emits matching M2 batches so the renderer can draw them correctly.
Adds WoweeModelLoader::toM2() and tryLoadByGamePath() to deduplicate the
identical conversion code that lived in editor_viewport for both objects
and NPCs. Cuts ~70 lines of duplicated logic and makes WOM->M2 reusable
across the codebase.
Upgrades WOM from geometry-only (WOM1) to fully animated (WOM2):
- WOM2 magic (0x324D4F57) for animated models, WOM1 for static
- Vertex extended: +boneWeights[4] +boneIndices[4] (40 bytes vs 32)
- Bone data: keyBoneId, parentBone, pivot, flags per bone
- Animation data: per-sequence per-bone keyframes with translation,
rotation (quaternion), scale at millisecond timestamps
- fromM2() now preserves all skeletal data: bone hierarchy, weights,
and per-sequence keyframes from M2 animation tracks
- Backward compatible: WOM1 files load without bone data (32-byte
vertices read and padded with default bone weights)
- FORMAT_SPEC.md updated with WOM2 binary layout
