A Quantized-Diffusion Model for Rendering SKIN and Translucent Materials

[Bigguns]

ho my god I was missing the boat! lol

I need to try that shader!!

I just read it all, That's great news vlado.

Yes a new interface would be awesome, plus a backscattering layer too, so 3 layers of sss, a la mentalray. I played with the new sss2 in mentalray, they have a R G B value for the radius, so you can play with it, so the more red you add, the more red you got into the shadow ,etc, if you want some greenish into the second sss layer you can, and it work well and give great control. Can you add something like that too?

here's an exemple : http://www.cgfeedback.com/cgfeedback...1&d=1355730937

ok both dont have the exact same placement of light and power, but you get the idea. dont look at the eyes of Vray it was an old eye set up and they are bad,( Vray can do amazing eyes too dno't worry haha). As you can see, I was able to replicate the skin feeling of Vray with Mr but with less greenish, I could aslo make more blueish if I wanted,etc. by tweaknig the RGB radius value of each R G B . so If I wanted the second layer of the front scatter to have a slighly more blueish tint in the shadow, I can put more B value and play with the weight of the sss layer,etc. so it give quite a lot of control.

Mr has a new upcoming sss2 shader, lightmap free (was about time, like this it will be faster and especillay more details will remain ni small render, I had to render in 2500 pixel height to keep decent level of details in the lightmap....) and will be layerable, so you can add let's say sss, 2 reflection layers(and dont use the crappy specular that come with it now lol ),etc, or 2 sss , 3 reflections layer ,etc so it will be very flexible.

That's the approach I would like too for Vray, bring a new interface, call it Vray skin shader, and give us a lot of control.
If some peoples are affraid of having more control, some are not. I prefer to have a lot of control and learn how it work then having it simple and less control. So I would suggest keep the sss2 as is, and do a skin material for us, the perfectionnist one so that way you keep compatibilities for older project and the peoples used to it.

Also did you find someting about the red bleeding sss occuring near close surface, like big wrinkles ?

What about adding sss bump ? can it help to retain even much details?

Adamlewis :in your exemple of the sillicon mask I found it's not green what we see but more a blueish greenish one and it's subtil... anyway it can be the lighting too as some pointed out etc.. and it's sillicon, not skin so.

anyway it's a great news to hear

[vlado]

Yes a new interface would be awesome, plus a backscattering layer too, so 3 layers of sss, a la mentalray. I played with the new sss2 in mentalray, they have a R G B value for the radius, so you can play with it, so the more red you add, the more red you got into the shadow ,etc

That's what you get with V-Ray too - the sss scatter color is a radius and obviously it has red/green/blue components. The point is, if you want to scatter more red into the shadows, there is less red remaining near the shadow edge (since the amount of light remains constant), so you get a cyan-ish tint in small details. This is of course, if one wants to stick to physical accuracy. But, as I said above, too much physics is probably not good in this particular case

here's an exemple : http://www.cgfeedback.com/cgfeedback...1&d=1355730937

Yes, I saw that already

and will be layerable, so you can add let's say sss, 2 reflection layers(and dont use the crappy specular that come with it now lol ),etc, or 2 sss , 3 reflections layer ,etc so it will be very flexible.

You mean like what V-Ray has been able to do for ages

That's the approach I would like too for Vray, bring a new interface, call it Vray skin shader, and give us a lot of control.

I think, that a specialized skin shader is indeed needed, so we'll see

Also did you find someting about the red bleeding sss occuring near close surface, like big wrinkles?

Well, I hope that the geometric sss sampling and the raytrace mode will help with this (as opposed to the prepass system that we have now - it's not the best thing for this task...)

What about adding sss bump ? can it help to retain even much details?

Possibly; but we need to implement raytraced sss first; trying to do this with the illumination map approach would be insane (it would need a very big density of the samples).

Adamlewis :in your exemple of the sillicon mask I found it's not green what we see but more a blueish greenish one and it's subtil...

Cyan-ish

Best regards,
Vlado

[peterguthrie]

good work Vlado!

[Bigguns]

Ho that sounds really good then Vlado, I think you know where you are going ,waiting for more :P My Daniel Craig model is crying for that new skin shader :P http://www.cgfeedback.com/cgfeedback...ead.php?t=4092

Edit : We need control over the backscattering :P

[Bigguns]

Please Vlado an opacity map slot for the sss2 pleaseee!

[anchovy]

Vlado-

Do you know if the same absorption curve bug shows up in the translucency of the VRayMtl when using hybrid or the other modes?

Thanks

-Michael

[vlado]

The translucency in the VRayMtl is not entirely physically accurate, plus it doesn't do multiple scattering. It does show discoloration of small details, but they are pushed to white always (see the attachment). I did write a perfectly accurate physical shader, and it matched the sss2 material quite closely.



Best regards,
Vlado

[anchovy]

Sweet! thanks for the info.

-Michael

[Mnietek2707]

I did write a perfectly accurate physical shader, and it matched the sss2 material quite closely.

Are you going to include this shader in future builds?

[vlado]

Are you going to include this shader in future builds?

No, it's very impractical - it takes a lot of time to clean up. I wrote it only so that I can verify the results from the sss2 shader. I can probably post it here, but I don't plan to officially support it.

Best regards,
Vlado

[3LP]

Yeah, show us some back office stuff

Stan

[adamlewis]

Hey Vlado, regarding the technique you mentioned for blending multiple SSS2 shaders together, I was wondering how you are accomplishing this exactly. I've been experimenting around with it and noticed that if I simply take two identical SSS2 shaders with only different scatter colors and radii and layer them in a VrayBlendMtl, the render result appears to only be an averaging of these two shaders, as opposed to an actual layering, evidenced by the fact that I can change the layer order and the result will remain the same (this is with a neutral gray swatch controlling the blend amount of course). This behavior isn't really surprising though, since this is pretty much the way a VrayBlendMtl behaves with any material if you're just using a swatch to control the blend amount, but I had the erroneous impression from your post that the VrayBlendMtl treats SSS shaders differently.

So the question is, what are doing to control the blend amount, so it blends within the volume of an object as opposed to merely mixing the scatter amounts of the different shaders, to create a proper layering effect. I see that, for example, the falloff map in Max has some options for blending based on distance etc. but nothing that seems to quite fit the roll needed in this case.

[vlado]

This behavior isn't really surprising though, since this is pretty much the way a VrayBlendMtl behaves with any material if you're just using a swatch to control the blend amount, but I had the erroneous impression from your post that the VrayBlendMtl treats SSS shaders differently.

You can imagine it like blending of multiple diffusion profiles into one profile, which would lead to the same results. Is it physically accurate? I.e. is there an actual physical material that corresponds to that combined diffusion profile? In most cases, probably not. Does it behave like you would expect? In most cases, yes. You can say the exact same thing about skin shaders found in other renderers.

Best regards,
Vlado

[anchovy]

Vlado-

I just retested this scene file with the current nightly build 2.55.01 nightly 051714. Looks like the scattering bug may have worked its way out of the fixes? Im not seeing the uniform color attenuation on the RGB like your test image.



Yours went far more yellow with very little perceptible green in it.

Any ideas?

Thanks

-Michael

[vlado]

I had to revert that back; what I did was not physically accurate and I could not figure out how to get the raytraced SSS to converge to the same result. You'll have to play with the colors a bit to get what you need.

Best regards,
Vlado