Subsurface scattering

Started by Dune, October 11, 2019, 10:39:45 AM

Previous topic - Next topic

KyL

looks quite tasty actually!

"Updated butter in pan." This is the kind of thing that always makes me laugh :)

Dune

Well, I now updated the whole cooking area :P  But perhaps this egg hasn't much to do with SSS, as I rendered this in an older version (in PT mode) and there wasn't much difference. Furthermore, I still can't get rid of the grain in the egg, whatever settings I change, and in PT the shadow is really dark; can't change that either.
So I will let the egg cool off a bit, before attempting another SSS breakfast.

BTW. I put the egg on a disc and lowered planet a bit here.

Dune

Tried it on a person. Glass shader before default shader with bump and textures. That doesn't work in defer all mode. It does in PT (best visible in nose), but with vitiligo >:( on some parts of the skin. It bleaches the skin considerably, so in the complete PT render I decreased default color from 0.5 to 0.3. Still rather light.
Other duties call, so I'll leave this to be crystalized a bit....

KyL

Yes, it's not working that great on humans. I tried a bit and came up with not-so-great results:
tgout-014 03m18s v4.4.38.1.jpg

The whole setup is totally physically inaccurate, and the actual "scatter" we see comes more from rough refractions than anything else... I think we will need a proper skin shader, or a separate section in the default shader to really have convincing skin.

Dune

Yes, actually I'm glad you couldn't get proper results either (so it's not my fault :P  ) And yes, I can't imagine we have to add glass shaders to everything we need to have some subsurface scattering to. Would be great if that could be added in a default shader indeed.

Would you mind sharing your settings for the buddha? Als AA and such? I seem to get quite some white grain in PT renders using the new glass.

KyL

Sure!

Glass shader:
-Decay Distance 0.1
-Volume density 2
-Caustics intensity: 0

Render node:
-detail 0.5
-AA 6, default robust adaptive sampler (1/64 with a 0.05 treshold)
-max paths: 64


Disabling the caustics and increasing the max paths should give you much cleaner/faster result than just increasing the AA in this kind of scenario!

Dune


WAS

#22
Quote from: KyL on October 16, 2019, 11:17:57 AMYes, it's not working that great on humans. I tried a bit and came up with not-so-great results:
tgout-014 03m18s v4.4.38.1.jpg

The whole setup is totally physically inaccurate, and the actual "scatter" we see comes more from rough refractions than anything else... I think we will need a proper skin shader, or a separate section in the default shader to really have convincing skin.

What are you using as a basis for comparison? I've seen this model used before, and when I see SSS demos, those are highly inaccurate and exaggerated for effect (like in games).



vs



It's possible your reflections are just too high, and masking some effect (he's in the open with bright ambiance in 360 degrees), because the effect on yours is more accurate than other peoples "true" results which glow in the lobe more than actual thin cartilage which is inaccurate and reversed on the same model.

KyL

#23
Wow, those examples are extreme! :o

Actually I used this model for years as a benchmark for skin shading. It was one of the early head scans with proper displacement and reflectance captured.
Here is how it looked in Arnold, back in 2012:
visage_arnoldRender_44min.jpg

We're far from it!
But keep in mind we are using a *glass* shader, which is meant to reproduce a completely different phenomenon.

What's funny is that it is now really easy to achieve a mix between transmission and subsurface in Terragen, something that can be quite tricky in other renderers!

Quote from: Dune on October 16, 2019, 12:15:50 PMThanks!
You're welcome. Sorry to pollute your thread with all those heads  :)

Oshyan

Dune, Defer All doesn't do the same subsurface scattering that PT does. Just so you know. 

- Oshyan

Dune

I know, Oshyan. Thanks. It was merely to show the difference between skin darkness. I think it's better suited for models without bump and additional shaders, but it needs more experimentation.

WAS

#26
Is it possible that mesh depth isn't correctly being interpreted for SSS? Looking at Dune's examples, and Kyls, some areas have a apparent SSS effect, and other areas, at the same depths (and light exposure, or similar) randomly don't, creating a blotchy "SSS" effect across the model.

Hannes larger, smoother object seems to give better results. Additionally I wonder how it would look on native TG assets. Glaciers, waves, crystal, etc. I have a lot of crystal type shaders to test eventually. We'll see.

Matt

KyL and Dune, what are your Micropoly Detail and Ray Detail Multiplier? Very short rays will fall back to the "surface point approximation" to avoid artifacts on displaceable surfaces, but at the moment this also applies to non-displaceable objects. So the Micropoly detail matters (and so does Ray Detail Multiplier if you change that from the default).
Just because milk is white doesn't mean that clouds are made of milk.

KyL

Quote from: Matt on October 17, 2019, 02:39:44 PMVery short rays will fall back to the "surface point approximation" to avoid artifacts on displaceable surfaces, but at the moment this also applies to non-displaceable objects

Ooooh! Interesting.
My tests were using 0.1 detail and 0.25 detail multiplier. I'll look into it.

Matt

OK, try it with micropoly detail 1 and an exaggerated decay distance. In the mean time I will try to make non-displaceable surfaces independent of the detail settings.
Just because milk is white doesn't mean that clouds are made of milk.