Curve/mesh shading discrepancies

[Dan Andersen]

Hello

I get some very different results rendering a curve and the same curve pre-meshed as a cylinder. Both are imported via the VRayMesh node.

Their normals appear identical, and they generally behave the same when rendered with a VRayMtl applied (diffuse, reflection and refraction).

But, as soon as I apply a hair material (VRayHairNextMtl for example) they behave differently. The curve seems correct, while the mesh goes very dark and dull.

Are there special mechanics in play for curves, that some shaders rely on? And if yes, is there a way of emulating that with meshes, so they can be rendered correctly with the VRayHairNextMtl?

Thanks,
Dan

[Dimitrina.Milanova]

Hello there,

I will have further look into this problem, however can you please attach a scene file and/or video of how the problem is reproducing.
Also what are you specs (OS, Maya version and V-Ray version)?

Best Regards,
Dimitrina

[Dan Andersen]

Hi Dimitrina, thanks for your reply. I can't attach any sample files right now, since I don't have the required permissions to upload files from from our internal network. I'll try and get this sorted out.

specs:
Maya 2018.5
VRay 4.3 (build #29690. I tested in a few different 4.x builds though, with the same results)
Ubuntu 16.04

[Dan Andersen]

Dimitrina.Milanova I sent you a PM with a scene showing the issue

[kalin.milenkov]

Hello,

Indeed the hair materials are specifically designed to work with the special class of hair geometry objects (which at some point resolve to a collection of curves). Even though the curves/hair strands appear as cylinders, the hair materials (VRayHairNextMtl in particular) model part of the light cylinder interaction implicitly. To do this they need some additional information such as the curve/hair strand tangent direction which is only from the hair geometry objects. On the other hand the model will ignore some of the apparent geometric hit information and use its own as predicted by the model it's based on (an example is the hit normal).
As consequence it is not currently possible to emulate the results from curves/hair with pre-meshed cylinders.

[Dan Andersen]

Hi kalin.milenkov , and thanks for the information - that's really helpful.

From this "VRayHairNextMtl in particular) model part of the light cylinder interaction implicitly", I gather the fiber shading model in VRayHairNextMtl probably assumes a perfectly circular strand cross-section, and wouldn't take into account rectangular or elliptical cross-sections, for example? (if we could actually create that)