Metallic IBL Reflection

[dgovil]

The USD Preview Surface shader setup only reflected IBLs (represented by an Arnold SkyDomeLight with a texture) if the specular color was not black. It also did this when the specular workflow was not enabled.

This PR fixes that (and addresses my needs in #933), so that metallic workflows also are capable of reflecting the IBL.

I also took the liberty of removing the autogeneration comments because, as far as we could discern, these are no longer direct results of the auto generation.

[dgovil]

@mattyjams and @JGamache-autodesk , the logic I'm applying here is pretty simplistic. I'm not sure if the metallic+roughness logic that is in usdPreviewSurfaceLighting.xml should also be represented here or if this is sufficient, since it comes after that particular fragment.

For reference, this is the code from the lighting fragment.

if (!useSpecularWorkflow) {
            float R = (1.0 - ior) / (1.0 + ior);
            float3 specColor = mix(float3(1.0, 1.0, 1.0), diffuseColor, metallic);
            F0  = R * R * specColor;
            F90 = specColor;

            // For metallic workflows, pure metals have no diffuse
            d *= 1.0 - metallic;
        }

At the very least, it allows for a basic reflection estimation in the viewport now

[dgovil]

We've tried to adapt the specular formula into the combiner with some modifications as we believe pertains to reflectivity.
@ysiewappl and I think this should be "close" but without a reference renderer to compare against exactly, we're unsure and would like someone more well versed in the UsdPreviewSurface like @mattyjams to verify when they can.
We are trying to make sure however that we do not modify the unlit mode in any undue way with our changes.

Below you can see a comparison of the before and after of the test sphere with our test IBL, with before and afters of this PR. I include the lit and unlit versions to discern how much things differ. Lit is forward/colorful, unlit is behind and greyscale since the sphere itself has no color. The shader setup in Maya is that the sphere is a light grey, with a black/white checkerboard in the specular/metalness parameters. The IBL is a validation room with a checkerboard pattern and bright text.

Similarly here is a comparison with the gramaphone asset in the Apple USDZ gallery. Again, each section includes the lit and unlit version of each asset. The IBL is a basic HDR room. I believe however the specularity of the metallic workflow should be tinted by the diffuse color, but I am not sure how to integrate that into this specific fragment.

In both images you can see that the unlit versions remain unaffected as we hope.
Prior to this PR, you can see that the metallic workflow has no IBL reflection since the shader fragments assumed only specular would reflect.
The luminance of non metallic parts remains consistent between the metallic and specular workflow as can be seen by the gramophone (Edit: the previous commit had a slight lift, but I fixed that and updated this comment and the images)

Note: I have not attempted to update the clear coat logic that @JGamache-autodesk mentioned yet. Perhaps that should be left to a secondary PR, once this one is sorted.

[JGamache-autodesk]

Already looks a lot better for metallic workflows. I have a small thing that might help if you want to try it, otherwise the code is a big step in the right direction.

[JGamache-autodesk]

There is a nice discussion on metallic looks here if you are interested: PixarAnimationStudios/OpenUSD#1174
Using the provided Damaged_Helmet_usda.zip asset in Maya with USD/pxr/imaging/hdx/textures/StinsonBeach.exr as the environment file should allow comparing the Maya shader against usdView.

[dgovil]

Thanks for that reference and the pointers, they help clarify quite a bit. I'll iterate a bit more internally and update the PR accordingly.

[dgovil]

I think your suggestions really helped.

Here's a new before and after.
The specular workflow remains unchanged.
Metallic now modulates its specular color by F0/F90 as a function of diffuse color, which means that specular highlights appropriately pick up the nature of the base color for metallic elements.

Let me know if you think there's any other changes I should try here, but otherwise I feel like this is in a good state for me personally.

[JGamache-autodesk]

Looks nice indeed. Thanks.
Does not require PF since all changes are in non-compiler VP2 fragment XML files.

[kxl-adsk]

Looking great indeed. Great job!
How about we get an image base test to make sure results of this hard work are not lost? Here is one example.

[dgovil]

@kxl-adsk I'll look into that however I had some questions:

1. The PF doesn't support Arnold which is required for the IBL map. I could spoof it with just colored lights though? Or I could write it to use the Arnold SkyDome and it could be disabled for the PF here.
2. If we do use the skydome light for the test, is there a specific IBL/HDR etc that your project prefers to use? The gramophone asset is MIT, or the sphere would be a new contribution so either would be fine to add to the test data set. However I do not have a license compatible IBL to add here as part of the test.

[kxl-adsk]

We are adding support for interactive tests in PF, but we won't have the ability to add them with additional dependencies (like MtoA). So Arnold SkyDome is not something we can use in tests.

@JGamache-autodesk do you have any recommendations on what would make sense to do in a test without Arnold dependency?

[JGamache-autodesk]

A double metallic/roughness ramp with two directional lights would probably be enough to make sure this does not regress:

from maya import cmds

for i in range(5):
	for j in range(5):
		sphere_xform = cmds.polySphere()[0]
		cmds.move( 2.5*(i-2), 2.5*(j-2), 0, sphere_xform, absolute=True )

		material_node = cmds.shadingNode("usdPreviewSurface", asShader=True)
		material_sg = cmds.sets(renderable=True, noSurfaceShader=True, empty=True, name=material_node+"SG")
		cmds.connectAttr(material_node+".outColor", material_sg+".surfaceShader", force=True)
		cmds.sets(sphere_xform, e=True, forceElement=material_sg)

		cmds.setAttr(material_node + ".diffuseColor", 0.944, 0.776, 0.373, type="double3")
		cmds.setAttr(material_node + ".useSpecularWorkflow", False)
		cmds.setAttr(material_node + ".metallic", i * 0.25)
		roughness = j * 0.25
		if j == 0:
			roughness = 0.001
		cmds.setAttr(material_node + ".roughness", roughness)

With a sky blue directional light rotated -45 in X and a yellow directional light rotated at +45 in X you get this incredibly ugly looking scene that yet makes sure all the metallic features are there. Of course it looks better with an IBL:


Without the IBL, there is no way to know that this is a gold material.

[dgovil]

Thanks! I'll get that set up as part of this PR.

[dgovil]

@kxl-adsk I've added a test to this PR. I also made some modifications to the utility libraries to handle changing cameras and differing color management environments (it was erroring in our environment where we had recently enabled color management workflows)