Odd issue with IBL moving with skinned mesh

Dealing with this odd issue when exporting gilt’s from Maya for use in three js. Has anyone come across this?
Screenshot 2023-11-09 at 6.45.47 AM
I’ve got 3 spheres. One rotated with an xform another rotated and xforms frozen and the third rotated with a joint. As you can see the joint rotated sphere seems to be reflecting the IBL and then taking it along for the ride in deformation. Has anyone come across this before?


  1. Which version of three are you using?
  2. Could you share the code or how the IBL is applied?
  3. Would it be possible to share the GLTF?

Yeah, I’ll share it. The screenshot is through VSCode’s gltf viewer. Which seems to use r153.I also see the same effect on https://gltf-viewer.donmccurdy.com which uses r158.

spheres_foribl.zip (24.2 KB)

Suspicion is def the gltf and some issue vertex normals, but I also lean toward some order of operations issue when sampling the IBL?

Seems like Maya GLTF exporter messes up the normals a little - I’m not sure how much maintenance it’s getting, if you’re using this one, the last commit was in 2021 :smiling_face_with_tear:

Just to follow up, it did end up being an issue with our exporter where it sampled vtx normals from an initial position of the deformed mesh vs sampling them from the intermediate mesh.
Thanks @mjurczyk for helping confirm the suspected normals issue!