Author Topic: [RESOLVED] Substance Designer (10.1.1): Yebis vs. IRay vs. Corona Renderer  (Read 1522 times)

Why do the Yebis and Iray renders look so much different than what is expected based on the render previews for a material as they appear on Substance Source.

See below for comparisons
Last Edit: May 13, 2020, 11:25:47 am

cross-post from Substance Forum » Substance Integrations » Autodesk 3ds Max


After rendering basically every permutation of the material using Substance renderers (Yebis, Iray) in various Substance material authoring / texturing softwares, I think I have to backpedal on my earlier assertions, because it appears I've misunderstood or miscalibrated Substance Designer.

Looking at the various renders (displacement exaggerated for effect) it appears that Substance Painter Iray render and Corona (tonemapped) render are the most visually similar to the preview renders from Substance Source. 

The Substance Painter Yebis and Iray renders differ significantly in terms of the normal map response - AO response? It looks like the self-shadowing is completely different, perhaps because the shadows are being calculated from the normal map in Yebis but from the geo with Iray?

---

substance source preview

vpq_substance_source_preview_1


vpq_substance_source_preview_2


---

vpq_corona_linear


vpq_corona_tonemap_5.0cm


vpq_corona_tonemap_1.0cm


vpq_substance_designer_iray_cw


vpq_substance_designer_iray_default


vpq_substance_designer_viewport_cw


vpq_substance_designer_viewport_default


vpq_substance_painter_iray


vpq_substance_painter_viewport


vpq_substance_player

Last Edit: May 05, 2020, 04:18:20 am

Just to reiterate the question:

Why do the Substance Designer previews, both Yebis and Iray, look so different from the Yebis and Iray previews *of the same material* in Substance Painter, and later in Corona Renderer?

Just want to bump this thread for some feedback - I've tried to document the issue as clearly as possible.

I'm using SD to replicate physical samples of material, so all of the surface properties of the material need to be simulated as accurately as possible, including base color, glossiness, normal, height, etc.

Of course, if what is rendered in the Yebis / Iray preview is not an accurate representation of how the surface will be rendered in a external renderer (i.e. Corona), then that is an issue.

It looks like the lighting is not correctly computed in SD, maybe because the shader does not match the node output, it's not clear. What outputs do you have in your graph ? And what shader is selected in the OpenGL renderer ?
Product Manager - Allegorithmic

Also: when using Iray, the surface normal is recomputed based on the displacement intensity, it's not the case in OpenGL.
So when using Iray the mesh has twice the normal information and the shading will look "wrong" / more contrasted.

There is no way to disable this behavior in Iray (nor in most rytraced/pathtraced renderers..), so you should probably set the normal intensity to 0 or at least lower it.
Product Manager - Allegorithmic

I conducted some more tests and it seems a lot of the differences were being driven by the HDRI used for testing.

If we compare the same Substance using the Tomoco Studio HDR in the IRay renderer, we see that the glossiness / roughness is MUCH more accurate with respect to the preview on Substance Source. However, the OpenGL render is still misleading in that respect - there is not strong parity between the glossiness / roughness response between the Iray and OpenGl renders and the Substance Source preview.

For example, in the renders attached below, the material looks soft, less glossy in the OpenGl render, whereas In the Iray render, the material looks more tense, more glossy.

Obviously OpenGL is far faster to iterate when building a Substance, so it is a little bit of a drawback if users that ultimately intend to render a Substance in an external raytracer / pathtracer must default to Iray to ensure accurate material response - at least there is a solution.

If there were a way to achieve parity between OpenGL and Iray renders, at least in terms of material response, that could be beneficial to all users.

What do you think?

Substance Source (preview)


Iray


OpenGL
Last Edit: May 12, 2020, 10:49:22 pm

In your preferences, in the 3D view section, you have an option to change the sample count for the OpenGL renderer. What is the current value on your machine? Do you see any changes if you increase it (like 128 for example).

Adjusting the OpenGL shader sample count improves the visual quality tremendously - thank you so much for your suggestion!

I was not aware of this sampling preference and adjusting the sampling to 256 (max) really helps bring the surface representation on par with Iray in terms of overall appearance. Thanks again @nicolas.cathaud, I really appreciate it!

OpenGL (sampling - 256)
Last Edit: May 13, 2020, 11:34:20 am

@nicolas.cathaud, @Nicolas Wirrmann

Is there a similar feature in Substance Painter - the visual disparity exists in that software as well? Of course there are bitmap filtering options, but is there an OpenGL shader sampling setting as in Substance Designer?