Efficient Skin Scattering For Cartoon Like Characters

[cybernoid]

Hi,

I see lots of fantastic complicated scattering materials for realistic skin on characters in other forum posts. But I am after something that is more efficient for non-realistic characters. Something like 'how to train your dragon'/'paranorman'/Pixar kind of thing.

In small isolated tests so far we have tried a mix of a 'diffuse with a scattering medium' and glossy material, and just one specular material. So far it seems the single specular material option might be a better choice - it seems more efficient then the mix choice and a better result.

It's for rendering an animation so efficiency is a concern but we must have scattering incorporated in it - I tried some hacks with just a glossy material and a falloff map but wasn't happy with it.

Also with a specular material it seems more consistent between 'direct lighting' and 'pathtracing' although I can't remember how much of a difference we should expect to crop up - I am worried about minimising the chance of unexpected consistency between shots, and we are prepared to do some with different kernels etc to balance production speed against quality etc.

Also when it comes to things like finger nails and other regions where we need to mix materials I wanted to check that for efficiency overall is it best to split polygons into material groups to avoid mix materials, and isolate the blends with mix materials to fewer polygons? It logically seems so but I wanted to double check before we do that extensively!

Does anyone have prior recommendations for this kind of skin setup?

Thanks!

[cybernoid]

In the hope this helps promote some discussion on the best way to do this I have attached an image showing my current path tracing tests with a hand and one specular material only with scattering.

The tests don't currently use any maps on the hand and I tried different degrees of solid fill values either side of the scattering effect to see what maps could be authored to blend between for the result needed (plus bump/normal/roughness maps eventually with possibly some falloff maps effecting sheen).

In my next tests I will try and see if the fingernails can be pulled off(!) with the same specular material or if I need to mix a glossy one in there - in which case that probably means I should split the end of fingers off into a separate material group from the rest of the hand so it can use a mix material for efficiency?

Also does anyone know what the scale in scattering medium actually relates too?

[churumbela]

Hi!
I've been recently investigating on medium nodes.
I've found this link very useful:

http://render.otoy.com/forum/viewtopic.php?f=21&t=9380

[cybernoid]

Ah that's useful regarding the scale thank you

[churumbela]

You're welcome. That's a nice looking character you're working on.

[riggles]

Ah that's useful regarding the scale thank you

That's funny, I was left scratching my head still. Wold you care to explain the scale control in a bit more layman's terms? I kind of wish there was just a depth control in real-world units.

[cybernoid]

Thanks the character is getting there - and he's more than a hand Getting the materials to look good takes ages though of course...

[cybernoid]

I think that link does a reasonable job of explaining it:

"The unit for absorption and scattering is m^-1 (or 1/m). If you have a measured value in cm^-1 you can convert it to m^-1 by
multiplying it by 100. For example: suppose you want to enter 4.5 cm^-1. This equals 450 m^-1. You can enter this value by
setting the scale to 1000 and the absorption texture to 0.45."

Where it is saying how much is absorbed or scattered per distance step.

Take: "This equals 450 m^-1. You can enter this value by setting the scale to 1000 and the absorption texture to 0.45"

As: scale 1000 * absorption 0.45 = 450 m^-1

Which I believe would be effectively the same as: scale 1.0 * absorption 450 = 450 m^-1

In the nodes you can't set absorption/scattering as values higher than 1.0 so the scale acts as a multiplier so that you effectively can.

Hopefully I got that right

[Rico_uk]

I like what you've got there on the hands, I'm finding it very difficult to get something that works in multiple lighting situations and with different types of geometry. I find that sometimes with the absorption I need to put the inverted colour that I need, and it seems like trial and error with transmission, absorption and scattering to get something that works, and usually it will only work for that one scene. I think the fake SSS method using a mix of glossy and diffuse works best for me and is the most reliable and predictable.
If Octane had a hue and saturation control in the Colour Correction node it would be massively helpful to try different things, not just for SSS but for other things too.
I would say that adding a very little emission to the shader does help to achieve a more toony look, especially with a bit of falloff.
Would love to hear more on this topic from other users and what their results are.