#vmdl

Two-stage flow: Blender headless emits Y-up OBJ, then pure Python writes vmats/vmdls and a C# model catalog.

Forge delivery job errors out but the build stage's previewGlbUrl stays downloadable — recover the GLB through Blender into your standard delivery shape instead of re-spending.

CC0 kits import like any OBJ+vmdl; city kits share colormap.png, nature kits need per-color vmats, and facing is a per-model yaw guess.

RenderMeshFile alone compiles clean with zero physics — ModelCollider needs PhysicsMeshFile/PhysicsHullFile in the vmdl.

Blender +X becomes world −Y after OBJ import; yaw +90° faces +X — and Blender +Y rotation tips the +X edge down.

CapsuleCollider and ModelCollider ignore WorldScale; BoxCollider follows it — bake scale into import_scale instead.

PhysicsShapeList + PhysicsJointList in the vmdl + ModelPhysics toggle — no collapse clips, no SetBoneTransform.

Scenes never load raw OBJ/GLB — always wrap with a .vmdl and remap materials under both bare and .vmat names.

Model compiler rewrites upper_arm.L → upper_arm_L; physics KV3 that still says .L makes limb bodies unknown while the torso works.

Default −Z/Y axes, bake_space_transform, no leaf bones, X/Z bone axes, mesh FBX + per-clip armature FBX; scale via ScaleAndMirror 0.3937.

AI mesh generators (Tripo, Meshy, Rodin) deliver textured OBJ+vmdl — but modeldoc32 headers, ~1m normalize, corrupt textures, bare texture paths, and missing colliders will ERROR the asset until patched.