topic Re: Normals and Morph targets in FBX Forum

Normals and Morph targets

Anonymous — Mon, 26 Oct 2015 17:58:07 GMT

Hi.

I'm looking at an issue involving morph targets and normals. I just started investigating and I wanted to get a little more information about them before I start honing in on problem areas.

We have an asset that is using morph targets with explicit normals set. When importing, the normals appear to have been averaged out. If the morph target is removed in the content editor (3DS Max 2014) and the asset exported, things import as expected. In our import code, the case with morph targets is seen to have a normal mapping mode of eByControlPoint, where the case without morph targets comes in with eByPolygonVertex.

Is it intended that the normal mapping mode in this case be eByControlPoint? If so, is the morpher modifier/morph targets mechanism compatible with meshes that have explicit normals (smoothing groups), or is this a limitation? Are there any known issues, either with the SDK or Max 2014 exporter, related this use case?

Thanks,

-Derek

Re: Normals and Morph targets

Anonymous — Wed, 18 Nov 2015 23:04:15 GMT

Wanted to bump/update this.

It seems that the problem is on the max exporter side. If I have a simple cube with a morpher modifier, it will consider the whole mesh to be contained in a single smoothing group and will export with a normal MappingInformationType of PerVertice.

I think this behavior makes sense, though with additional information one could recompute appropriate normals for faces after a vertex has been interpolated towards its morph target. So, I'm still left wondering about smoothing group support. If I'm attempting to FBX-export a mesh with morph targets and smoothing groups from Max 2014, should the resulting FBX file necessarily replace the smoothing groups with a single smoothing group and use per-vertex normals?

I've been trying to look for documentation on the technical details for how this modifier is supposed to work and what its limitations are, but there's been no mention of this as a limitation so I suppose that it is supposed to work and is a bug? I'm currently using Max 2014 with the 2014 version of the FBX exporter. We will be upgrading to more recent FBX versions in the near future, so if this is a known fix in later versions of the SDK and/or Max's exporter plugin, that would be useful information. We also have Maya available, so if there are any known alternative processes that make use of Maya, that information would be much appreciated as well.

Sorry for the wall of text. Thanks for taking the time and reading, if you did. I'm a programmer and not an experienced Max user, so if there are issues/misunderstandings with any procedure I described, please feel free to let me know. 🙂

I've attached a basic example (vanilla_box.max):

The file contains one simple cube and a perturbed morph target.
The cube's smoothing groups match with the morph targets.
Three corner faces of the cube are in a single smoothing group, the other three are in distinct smoothing groups.

I attached an example of exporting with the morph modifier (vanilla_box_morpher_ascii.fbx) and without (vanilla_box_nomorpher_ascii.fbx).

Finally, a picture of my export settings:

Re: Normals and Morph targets

Anonymous — Fri, 08 Apr 2016 03:02:18 GMT

It looks like all the normals have been averaged out. It could be an exporter issue. Here are the normals for the box and the target.

Re: Normals and Morph targets

Anonymous — Fri, 08 Apr 2016 03:15:41 GMT

Hey Numerator.

Thanks for the photos.

I was originally looking for information on what to expect when you use the morpher modifier with smoothing groups. When I looked at an FBX export of this from Max2014, the normal type and normal information was different (depending on whether the modifier was present or not). In the modifer case, the normals were "by-control-point", giving an averaged/interpolated result.

The expectation from artists was that the more complicated scenario of maintaining both the smoothing groups and the morph modifier would occur.