New vray mtl SSS , pro and con's

[Bigguns]

Hi there, I been trying the new sss shader in vray,, tested it in and out, using brute force on both engine and at 24 bounces.

The things I don't like about this shader are these and in order of importance:

The shader itself look's flat, like if there was a highpass on the SSS , it's very strange, it may be ok for plastic SSS but definately not working for realistic , I had to stick to the Vrayalsurface because it show's more form in the SSS, the new vray SSS is so flat.

Second thing : the new vray SSS has a HUGE shadow shifting occuring compared to a blinn ( and it's a WAY bigger one then the vrayalsurface in directionnal mode ) .


The thin I like : I can see in some places it's more accurate ( and even more noticeable on high radius sss objects ) , but the flattening effect and the shadow shift are too big of a deal for me to use it on character, it's definately less realistic for skin then the alsruface.

So it has lot's of potential as a shader, but these 2 things needs to be address in order to be able to render convincing skin with that sss shader.

[Muhammed_Hamed]

Hi Bigguns Thanks for the feedback

The shader itself look's flat, like if there was a highpass on the SSS , it's very strange, it may be ok for plastic SSS but definately not working for realistic

Could you please share an example of this flat look? I don't know what you mean here
I've been using the new SSS in every scene since 5.1, dropped ALsurface and FastSSS2

This example here is simply not possible to achieve with ALsurface or FastSSS2
https://cdn.discordapp.com/attachmen...095370/001.jpg

And even before 5.1 I would use V-RayScatterVolume material for complex SSS setups. It had a similar RandomWalk implementation and allows for multiple scatters inside the volume(ALsurface is one single scatter and falls short of said quality)
https://cdn.discordapp.com/attachmen...mage_38448.jpg

the new vray SSS has a HUGE shadow shifting occuring compared to a blinn ( and it's a WAY bigger one then the vrayalsurface in directionnal mode )

Are you using Volumetric mode here?
Would be nice to see an example or more information on your setup.

Hi there, I been trying the new sss shader in vray,, tested it in and out, using brute force on both engine and at 24 bounces

Is there a reason for using BF/BF here?

Best,
Muhammed

[Bigguns]

Hi Bigguns Thanks for the feedback



Could you please share an example of this flat look? I don't know what you mean here
I've been using the new SSS in every scene since 5.1, dropped ALsurface and FastSSS2

This example here is simply not possible to achieve with ALsurface or FastSSS2
https://cdn.discordapp.com/attachmen...095370/001.jpg

And even before 5.1 I would use V-RayScatterVolume material for complex SSS setups. It had a similar RandomWalk implementation and allows for multiple scatters inside the volume(ALsurface is one single scatter and falls short of said quality)
https://cdn.discordapp.com/attachmen...mage_38448.jpg



Are you using Volumetric mode here?
Would be nice to see an example or more information on your setup.



Is there a reason for using BF/BF here?

Best,
Muhammed

You are not using it for skin ! I'm talking for skin, for skin it's way less good then Alsurface because of the con's I said. I will post images soon, I'll do some to show what I mean. If they can fix these 2 things, yes it would be better then alsurface. But for skin it's not.

[Nearl]

My issue with VRayMtl SSS (vray5 update 2.2) is the render times. 4+ hours vs 16 minutes for the AlShader (I only swapped the shader). The AlShader also has issues when the geometry become to thin and in-between multiple objects (you can unify the Raytraced SSS ID for all objects but I tend to get blow outs where they meet). I tried a simpler test... AlShader 4m21.5s, Fast SSS 2m34.8s, VRayMtl 9m48.2s



Any suggestions for optimal settings when using VRayMtl SSS?

[^Lele^]

If you could get me the scenes via email, then i could take a look and be more precise.
In general, the V-Ray material SSS is currently slower than the other approaches as it's *by far* the most accurate (true random walk).
It's quite sensitive to lights count, so if you need more speed, try and make the number of lights lower (f.e., baking what you can in a probe).
Further, the method is by nature sensitive to the inner geometry (f.e. the skull and muscles inside the head), whereas the other SSS models ignore that entirely.
This means that for long path settings the RW SSS may take longer with the same geo and shader setup than AlS would, but it would also produce different visual results (as visible in your images).

[Nearl]

Thought this was a cool test. Random walk with No underlying geometry, with underlying geometry and the underlying geometry. Just thought I would share.

[^Lele^]

Super cool!

[vlado]

Very cool tests! There are some improvements to the SSS in the VRayMtl material including speed improvements - as Lele mentioned, getting some scenes for testing the performance would be awesome.

Best regards,
Vlado

[Nearl]

Very cool tests! There are some improvements to the SSS in the VRayMtl material including speed improvements - as Lele mentioned, getting some scenes for testing the performance would be awesome.

Best regards,
Vlado

I can send you over my arm asset...my other examples were developed for clients. Should I email it to Lele or another address?

[^Lele^]

Me or Vlado is equally fine, consider all assets you send us under NDA.
And thanks SO much for them!

[Bigguns]

Me or Vlado is equally fine, consider all assets you send us under NDA.
And thanks SO much for them!

Hi Lele, I'm preparing 2 scenes for you about the alsurface problem it has and the new sss of the beta 6 vray, I'm testing the beta 6 right now. the new uniform mode is a lot better then the other one for skin, it's almost identical to the alsurface in diffuse mode with single sheet sss. I will give you 2 scene, one for the alpha not working in the sss of the vray mtl ( it never did work ) and the other one it is the 3 shader to compare. The alsurface in directionnal mode ( always single sheet sss, using the 3 sheets don't lead to realistic skin at all ) the diffusion mode of the alsruface and the new vray sss mtl in uniform mode. If you add textures to them you will see that the uniform sss mode of the vray mtl don't hold the color the same as alsurface do, and I think alsurface does it the right way. I'll show you an exemple also, I will join a few images. Is this email : the emanuele.lecchi@chaosgroup.com the right one to send you my confidential files ? thank you

[^Lele^]

It should actually be @chaos.com going forward (although chaosgroup.com would work as well for a while at least).
My bad: i should fix the signature, thanks for noticing.
Also many thanks for the files!

[Bigguns]

It should actually be @chaos.com going forward (although chaosgroup.com would work as well for a while at least).
My bad: i should fix the signature, thanks for noticing.
Also many thanks for the files!

Alright thank you. I'm still working on it, preparing many renders for comparaison. The best would be if we can speak on skype and I share you my screen to really explain better what I mean but I will try writing it in the email. Let me know if you can skype that would be good, I was skyping with Kalin when I was beta testing the hair shader and giving him feedback and suggestion about the glint and glitter and it did worked out great.

[Bigguns]

I'm doing extensive testing. Clearly the way the albedo map is displayed is wrong, the SSS interact with it changing it's color, it needs to be done like the Alsurface shader does it. Second thing is making the alpha work. When I render only the sss without the diffuse it look's pretty good( but with the dalbedo, it look's less good then Alsurface. Also it's quite slower then Alsurface, but once the albedo display problem fixed and the opacity working, I think maybe it will be better then Alsurface. more testing will be needed when we get there. so first one is alsurface in diffusion mode (single sss sheet ) then alsurface directionnal ( single sss sheet ) and the new vray sss in uniform mode.

[Bigguns]

It should actually be @chaos.com going forward (although chaosgroup.com would work as well for a while at least).
My bad: i should fix the signature, thanks for noticing.
Also many thanks for the files!

Hi Lele, I sended you an email from wetransfer for a scene and explanation about the problems I found in the shader. let me know if you have any questions