Shading improvement thoughts

[Ihno]

we are looking into better filtering for bump and normal maps and we've done some experiments, but things are more complicated when applied in practice than what the papers are suggesting.

Would that make creating microscraces as shown above (in the video with the cutlery) easier? Right now its almost impossible. Or is that a different problem?

That's a good question though, if you had to choose between a X% slower render and better energy preservation, which would you choose?

If its meant to be always on - 2.5% seems ok everything above 5% would be a problem for me I think.
If nessesary: The percentage would go up a bit if it's switchable.

[FLP]

If you had to choose between a X% slower render and better energy preservation, which would you choose?

I would choose what is consistent and predictable. If I can see/photograph it in real life I should see a corresponding function or result in vray. Render time can be solved by throwing more Ghz at it. Shader development takes man hrs which ultimately cost more.

[vlado]

For the energy preservation, I guess you're talking about the Heitz's Multiscatter model which solve the light interaction completely? What about energy compensation like in the links provided in the first post? It's probably less physically accurate but anyway, it's energy conservative, reciprocal and it provides really good results IMHO. The only drawback is that it needs to be precomputed and stored in a LUT. But I would be curious to see the performance hit of this, it could be totally worth it.

I meant the multiscatter model. I had, in fact, missed the other links - thanks so much for posting these; the approach there seems a lot easier to implement. Will definitely look into it.

Edit: Do we have to choose between the two? Enabling/Disabling the feature is not an option?

I would prefer if things worked better out of the box without additional options

Best regards,
Vlado

[kosso_olli]

I remember there was some kind of extended version of the stochastic flakes in development which allows prodedural generation of scratches, no?
Energy preservation would be great, we have this quite often on glossy parts. And there are lots of them on a car...
Also Oren-Nayar is highly welcome if it fixes the harsh shadow terminator issue and improves the falloff between light and shadow, especially under hard and contrasty lighting.

[vlado]

Also Oren-Nayar is highly welcome if it fixes the harsh shadow terminator issue and improves the falloff between light and shadow, especially under hard and contrasty lighting.

No, it doesn't do that, I think. These issues will need to be solved in another way.

Best regards,
Vlado

[CCOVIZ]

the approach there seems a lot easier to implement. Will definitely look into it.

Nice to read that! You should have a close look at the Stephen Hill blog post, he seems to fix some caveat of the initial presentation.

I would prefer if things worked better out of the box without additional options

Best regards,
Vlado

Totally understand that!

[kosso_olli]

No, it doesn't do that, I think. These issues will need to be solved in another way.

Best regards,
Vlado

Well, judging by the pictures posted on page 1 from V-Ray for Modo, apparently it does.

[vlado]

Well, judging by the pictures posted on page 1 from V-Ray for Modo, apparently it does.

Hmm. Interesting. You'll be able to check it out anyways

Best regards,
Vlado

[1funk]

vlado are there any plans to get diffuse roughness working on GPU?

[vlado]

vlado are there any plans to get diffuse roughness working on GPU?

Yup.

Best regards,
Vlado

[CCOVIZ]

This may help for the terminator problem :

https://link.springer.com/content/pd...-4427-2_12.pdf

[gaia]

Hello all this thead is really interesting !
what's news about that ?

[anchovy]

That's a good point about the diffuse roughness, it's been on my "to do" list to switch to the Oren-Nayar implementation since forever, it's probably time

Thanks for the other suggestions as well; we are looking into better filtering for bump and normal maps and we've done some experiments, but things are more complicated when applied in practice than what the papers are suggesting.

Better energy preservation for glossy materials is also on the "to do" list; the trick here is to do it without sacrificing performance. That's a good question though, if you had to choose between a X% slower render and better energy preservation, which would you choose? (Obviously depends on the X, but how much would be acceptable?) For Blender, someone mentioned about 2.5% slowdown, but it might be actually higher in V-Ray because the BRDF code is more critical to performance.

Best regards,
Vlado

Vlado-

Any more exploration in this area? Missing Oren-Nayar

Thanks

-Michael

[kosso_olli]

Indeed, did Oren-Nayar get implemented as rooughness model for the VrayMaterial in 3dsMax? Lost track of this in the changelogs, I have to admit.
I also recently asked Lele about the research of stochastic scratches feature. Was there any outcome?