Significant artificing in the reflection when hair and cloth shading models overlap in 5.5.4

There is significant aliasing artefacting on the cloth shading model when overlapping with the hair shading model.

  • This occurs at the point where they overlap there seems to be some kind aliasing effect.

Additional information:

  • This effect is strongest when cloth has a metal value of 1 and fades as the value is reduced to 0.
  • Is strongest with a roughness of 0.5 fading when the roughness approaches 0 or 1.
  • None of the other values on the cloth material matter.
  • None of the values on the hair seem to matter.
  • I haven’t observed it with any other material combination.
  • It doesn’t appear in any of the buffers that I can find.
  • It doesn’t mater if the hair is in front or behind.
  • It effects the reflection. If you turn off the reflections it goes away, but is doesn’t matter if it is screen space or lumen reflections.

This appears to be addressed in 5.6.

Our project will be upgrading to 5.6 so this is not an issue for us, but I am providing infomation for awareness and in case it shows up again.

Thanks for you help.

Steps to Reproduce
Create 2 materials:

  • Material 1 - Cloth
    • Cloth shading model
    • Metal = 1.0
    • Roughness = 0.5
  • Material 2 - Hair
    • Hair shading model
  • It is important to use the specific settings on the cloth, as a non metal cloth or a roughness of 0 or 1 does not display the artrfacting.

Place overlapping meshes in a scene.

  • This bug shows up most clearly:
    • Several broad shapes like spheres with the cloth material on
    • Several layered up planes with the hair material on are placed in front.

Observe significant artefacting on the mesh using the cloth shading model.

By placing a post process volume into the level and turning on and off reflections the artefacting appears and disappears.

Here is a video of the reflection being turned on and off.

Hi there,

Thanks for the bug report.

Since you do not require a workaround, I am closing this issue for now, as I can confirm that it has already been fixed in 5.6.

For future reference, if anybody requires a fix in 5.5, I will leave some notes here as reference to give a starting point on where to look for the fix / fixing commit.

First I will note that I was unable to reproduce this issue with the supplied steps. However, setting the ambient occlusion on the cloth material to 0 reliably reproduced the same artifacts (the metallic value was not required).

Seconds, the artifacting seems to first appear in the Lumen.ScreenProbeGather.RoughSpecularIndirect buffer (RenderDeferredLighting -> DiffuseIndirectAndAO -> LumenScreenProbeGather -> Integrate -> SimpleDiffuse/SupportAll)

[Image Removed]

The issue probably stems from an interaction with the Lumen.ScreenProbeGather.TileClassificationModes buffer, which is marking these tiles with different classifications

[Image Removed]

Regards,

Lance Chaney