PBR shader discussion

[bluebox]

Wow, the Oren-nayar looks so much better!

[Naxos]

Wow, the Oren-nayar looks so much better!

Sorry, i don't get it... In my opinion / habits, Lambertian looks better here...

Can you explain why i'm wrong ?

[bluebox]

Wow, the Oren-nayar looks so much better!

Sorry, i don't get it... In my opinion / habits, Lambertian looks better here...

Can you explain why i'm wrong ?

What do you want me to explain ? I staded my subjective opinion, you stated yours. IMHO oren-nayar looks way less plastic - obviously visible in the mud exaple.
Looks also more convincint at grazing angles. Sort of like the lambertian model had kind of a black halo on 0 angle.
Hard to tell the real difference without proper testing in various conditions tho.

[pokoy]

Oren Nayar is perfect for rough, dirty, dusty, non-shiny surfaces - mud, soil, stucco walls, concrete, rubber etc. It's something you can see when you look at the moon - it doesn't look round, it looks almost flat, because of the rough surface. Oren Nayar is quite old though, I believe there are newer, better shading models available now. Still, it would be great to have anything that helps with these surfaces.

[Juraj]

The most recent PBR shaders also adjusted roughness of diffuse BRDF automatically with specular BRDF.

So not like old Vray shader where Diffuse Roughness simply interpolated between Lambert (default 0) and Oren (1), and enabled monstrosities like fully rough diffuse + fully specular GGX response on top :- ).
Instead, setting the general roughness parameter to 0.4 example would adjust both BRDF of the diffuse and the specular part.

This sort of approach, not giving user the option to created artificial crap is the right way to go.

But I also believe simply adopting Oren-Nayar would be half solution at this time. There must be something new, better, more universal. The difference between Lamber and Oren-Nayar is too radical even when interpolated in-between.
But it's still better than having only Lambert.

[Fluss]

I'm a little dubitative while looking at the results shown above. I wouldn't expect that much of an impact from oren-nayar alone. Those are marmoset toolbag renders right?

The old Vray model was not even oren-nayar, it's a custom gamma-based model that looks like.... em... Oren-nayar is way better ! I wouldn't even consider the Disney one. Even if they manage to get the same roughness as GGX, it is not energy conservative...

The only issue is that Oren-nayar and GGX roughness don't match and as far as I know, there is no good conversion between those two. Some attempts have been made in the real-time space, but it does come with trades-off.

[LorenzoS]

Wow, the Oren-nayar looks so much better!

On my opinion Lambertian is much better

[JViz]

I'm a little dubitative while looking at the results shown above. I wouldn't expect that much of an impact from oren-nayar alone. Those are marmoset toolbag renders right?

The old Vray model was not even oren-nayar, it's a custom gamma-based model that looks like.... em... Oren-nayar is way better ! I wouldn't even consider the Disney one. Even if they manage to get the same roughness as GGX, it is not energy conservative...

The only issue is that Oren-nayar and GGX roughness don't match and as far as I know, there is no good conversion between those two. Some attempts have been made in the real-time space, but it does come with trades-off.

yes exactly, these at best are artistic pseudo-mathematical representations of Oren-Nayar, any google images search for Oren-Nayar would give many images comparing Oren-Nayar's implementation with physical models and then Lambert, it's very clear Oren-Nayar is closer to reality. if somebody doesn't like how a shader looks if it's closer to reality I think it's very clear where the problem is!

I admire your insight Fluss, I think we are at a point where theory is experiencing a schism, one side applies to glossy materials while the other applies to rough materials. this is not pretty. a pure shader is out of reach for now, the best we can hope for is a blend that will break at many instances. unify the forces!! lol the theory of everything anyone?

[bluebox]

Can't we adapt the way Arnold handles things ?
https://answers.arnoldrenderer.com/questions/2521/lambert-oren-nayar.html

Everyone would be happy being able to choose pure lambert, pure oren-nayar and anything in between.

My point asking about the shader development on the topic here was to start some sort of discussion with the Devs as to what people expect and that those more tech-guys could toss some knowledge that could be easily overlooked.
Can we hear the Dev side please ? What are you guys planning to include in the new shader ?

[JViz]

Can't we adapt the way Arnold handles things ?
https://answers.arnoldrenderer.com/questions/2521/lambert-oren-nayar.html

Everyone would be happy being able to choose pure lambert, pure oren-nayar and anything in between.

My point asking about the shader development on the topic here was to start some sort of discussion with the Devs as to what people expect and that those more tech-guys could toss some knowledge that could be easily overlooked.
Can we hear the Dev side please ? What are you guys planning to include in the new shader ?

the Trello page says Clear Coat, Sheen, and Advanced diffuse model.

[bluebox]

Can't we adapt the way Arnold handles things ?
https://answers.arnoldrenderer.com/questions/2521/lambert-oren-nayar.html

Everyone would be happy being able to choose pure lambert, pure oren-nayar and anything in between.

My point asking about the shader development on the topic here was to start some sort of discussion with the Devs as to what people expect and that those more tech-guys could toss some knowledge that could be easily overlooked.
Can we hear the Dev side please ? What are you guys planning to include in the new shader ?

the Trello page says Clear Coat, Sheen, and Advanced diffuse model.

Are you a part of the development team JViz ;) ? I am familiar with trello roadmap. Advanced diffuse model can be interpreted in various ways.

[Fluss]

I admire your insight Fluss, I think we are at a point where theory is experiencing a schism, one side applies to glossy materials while the other applies to rough materials. this is not pretty. a pure shader is out of reach, for now, the best we can hope for is a blend that will break at many instances. unify the forces!! lol the theory of everything anyone?

Well, there is really nothing to admire here, just sharing stuff I read and liked. Some will share the same point of view, others won't and it's totally fine that way. But you are right if the shader is being rewritten, it should integrate the last advances in the field. And once the hype around the Disney shader as fallen a bit (I was there too), I have to admit that it's not the best one to look at for the aforementioned reasons ( it's principled but not technically phisically accurate).



In that regard, the Arnold Autodesk standard surface is way better. The diffuse model is good, the sheen is way better (physically based ->http://www.aconty.com/pdf/s2017_pbs_imageworks_sheen.pdf , not like Disney one) and it has everything we need (additional specular lobe, thin-walled surfaces for both diffuse and specular, thin film etc...).

What's more it is shared as a standard for material exchange, OSL and materialX available here: https://github.com/Autodesk/standard-surface/tree/master/reference.

add to that :

_ multiple scattering for every micro-faceted lobes (links some posts above)
_ real Shadow terminator Fix (micro-facets based shadowing function) -> https://link.springer.com/content/pdf/10.1007%2F978-1-4842-4427-2_12.pdf
_ micro-facets based normal maps -> https://cg.ivd.kit.edu/publications/2017/normalmaps/normalmap.pdf
_ same for bump maps ("bump to roughness")

and you get a pretty solid shader.

[JViz]

Can't we adapt the way Arnold handles things ?
https://answers.arnoldrenderer.com/questions/2521/lambert-oren-nayar.html

Everyone would be happy being able to choose pure lambert, pure oren-nayar and anything in between.

My point asking about the shader development on the topic here was to start some sort of discussion with the Devs as to what people expect and that those more tech-guys could toss some knowledge that could be easily overlooked.
Can we hear the Dev side please ? What are you guys planning to include in the new shader ?

the Trello page says Clear Coat, Sheen, and Advanced diffuse model.

Are you a part of the development team JViz ;) ? I am familiar with trello roadmap. Advanced diffuse model can be interpreted in various ways.

I'm not part of the development team, should I therefore sh**up? ;) cheers bro

[JViz]

I admire your insight Fluss, I think we are at a point where theory is experiencing a schism, one side applies to glossy materials while the other applies to rough materials. this is not pretty. a pure shader is out of reach, for now, the best we can hope for is a blend that will break at many instances. unify the forces!! lol the theory of everything anyone?

Well, there is really nothing to admire here, just sharing stuff I read and liked. Some will share the same point of view, others won't and it's totally fine that way. But you are right if the shader is being rewritten, it should integrate the last advances in the field. And once the hype around the Disney shader as fallen a bit (I was there too), I have to admit that it's not the best one to look at for the aforementioned reasons ( it's principled but not technically phisically accurate).



In that regard, the Arnold Autodesk standard surface is way better. The diffuse model is good, the sheen is way better (physically based ->http://www.aconty.com/pdf/s2017_pbs_imageworks_sheen.pdf , not like Disney one) and it has everything we need (additional specular lobe, thin-walled surfaces for both diffuse and specular, thin film etc...).

What's more it is shared as a standard for material exchange, OSL and materialX available here: https://github.com/Autodesk/standard-surface/tree/master/reference.

add to that :

_ multiple scattering for every micro-faceted lobes (links some posts above)
_ real Shadow terminator Fix (micro-facets based shadowing function) -> https://link.springer.com/content/pdf/10.1007%2F978-1-4842-4427-2_12.pdf
_ micro-facets based normal maps -> https://cg.ivd.kit.edu/publications/2017/normalmaps/normalmap.pdf
_ same for bump maps ("bump to roughness")

and you get a pretty solid shader.

thanks for the links Fluss! I'll make sure to go through it all, Arnold's shader is quite solid it seems. the dev team is not being very transparent so far and that's understandable because it's very probable that they don't yet know what they can achieve within the time frame of V6 release.

[Fluss]

Looks like things starts to move, Clarisse will embed the Autodesk standard surface shader and now otoy is doing so.