Wikihuman and Chaos labs

[sharktacos]

So Chaos group now has Chaos Labs, and one of it's projects is the Wikihuman project. I'm wondering if someone at Chaos could comment on the approach to the skin shader used there, which differs both from SSS2 and VraySkin?


Here's what I could piece together looking at the Vray shaders in provided Maya file:

The provided Vray shader mixes 2 phong spec lobes (60% + 40%). This basically seems identical to the VraySkin material which also mixes two phong spec lobes. The way that the Fresnel is applied might be slightly different.

It also adds the subsurface from an SSS2 together with grey Lambertian diffuse from a VrayMtl (to get single scatter). This approach to the SSS/diffuse mix differs from SSS2 where the diffuse input does not add to the SSS, but replaces it (on SSS2 diffuse amount 0.5 means that you get 50% diffuse and 50% SSS).

Additionally, the approach taken here is akin to the Alsurface material (a free shader written for Arnold by Anders Langlands) which takes a single diffuse map for the skin resulting in a WYSIWYG approach, in contrast to VraySkin which uses multiple maps (diffuse, shallow, medium, deep) which often need to be pretty non-intuitive colors like green for the diffuse and mustard yellow for the medium scatter.

I'd love to hear some discussion about the thinking behind this approach to skin shading.

[FSGFX]

bigguns is allready asking the same here...
http://forums.chaosgroup.com/showthr...leaseeee/page4

[sharktacos]

Thanks, that's an interesting thread. It's mainly about the Arnold shader tho. I think the Wikihuman project deserves a thread of it's own since it's a project Chaos is specifically involved in

[vlado]

I'm not quite sure what's your question? Since there is only one scatter map, it's more natural to use the sss2 shader than the VRaySkin one. The alSurface shader is a lot closer to the sss2 shader, only it has 9 subsurface bands instead of just 3 (i.e. it has 3 scatter color parameters instead of just the one in sss2).

The VRaySkin shader was created solely for those people that cried about other renderers having it...

Best regards,
Vlado

[sharktacos]

I'm not quite sure what's your question? Since there is only one scatter map, it's more natural to use the sss2 shader than the VRaySkin one. The alSurface shader is a lot closer to the sss2 shader, only it has 9 subsurface bands instead of just 3 (i.e. it has 3 scatter color parameters instead of just the one in sss2).

The VRaySkin shader was created solely for those people that cried about other renderers having it...

Best regards,
Vlado

I was just curious to hear the thinking behind the blends in the shader hierarchy. For example using diffuse from a VrayMtl added to an SSS2, as opposed to just having the diffuse from the SSS2. Basically I just was hoping to hear the thinking behind how the shader is put together here since it seemed interesting.

[jasonhuang1115]

IIRC, Mike (Fxguide) did an interview with Chris Nichols that was released around last year FMX as either a FXguide tv or one of Mike's background fundamental video where Chris broke down the thought process and details of the said shader. Dig around and you should be able to find it online.

[joconnell]

I was just curious to hear the thinking behind the blends in the shader hierarchy. For example using diffuse from a VrayMtl added to an SSS2, as opposed to just having the diffuse from the SSS2. Basically I just was hoping to hear the thinking behind how the shader is put together here since it seemed interesting.

Pretty much just better control over the falloff of reflections and blending - there isn't enough fresnel control in the sss2 at the minute so folks were using the regular vray material for the thin paper layer of the skin itself and the reflections, then the sss2 for all the sub surface lighting bits.

[sharktacos]

One thing that I wish was discussed more is the acquisition of the maps... assuming one is not able to have Paul Debevec scan them for you

[joconnell]

It's covered here - http://gl.ict.usc.edu/Research/DigitalEmily/

[sharktacos]

It's covered here - http://gl.ict.usc.edu/Research/DigitalEmily/

I think this is for Emily1 rather than Emily2. For example, Emily2 uses a single scatter map which is derived by extracting the specular from the cross-polarized difference map which contains both specular reluctance and single scatter. I do not think the Emily1 page above, nor the PDF paper on it, cover how this is map is generated.

I did find it covered on page 4 of this paper:http://gl.ict.usc.edu/Research/LFR/f...nce_sa2008.pdf

but honestly, I don't think I could reproduce without special software. The paper is written rather cryptically, so it's possible I am just not getting it, but at any rate I am not getting it. Are you able to make sense of it?