Adding random walk SSS to alsurface please

[DaveKW]

It seems to work differently with the scattervolume as the base. As there are no diffuse, reflection or refraction properties, sitting a fully refractive shader on top at 100% doesn't override the base shader as it would normally.

[MANUEL_MOUSIOL]

since this is in the wishlist, I will just hijack the thread and ask when scattervolume will be available for GPU?
thanks and sorry for the interruption.

[Bigguns]

this volumes shader don't work well on skin at all . The problem is in the valley, it look's very dark and blueish ( or greish depending of the color choosed ) ... It look's good when the sss effect is strong across all of the object ( wich is not the case skin ) like maybe a very translucent something lol. Otherwise, it's like if there is no bouces happening in the valley's and it becomes greyish.. I did extensive test of this shader, either in vray or corna.. got better result in corona, but still it had the same prbolem ( also, very hard to get a decent diffuse color that match the color of the map you put in, was looking darker etc... ) in arnold it work's way better, especially the new randomwalk2 .

[jasonhuang1115]

this volumes shader don't work well on skin at all . The problem is in the valley, it look's very dark and blueish ( or greish depending of the color choosed ) ... It look's good when the sss effect is strong across all of the object ( wich is not the case skin ) like maybe a very translucent something lol. Otherwise, it's like if there is no bouces happening in the valley's and it becomes greyish.. I did extensive test of this shader, either in vray or corna.. got better result in corona, but still it had the same prbolem ( also, very hard to get a decent diffuse color that match the color of the map you put in, was looking darker etc... ) in arnold it work's way better, especially the new randomwalk2 .

Bigguns I was interested in trying out the scattervolume on skin but haven't got to it. Since you have done extensive test, any test renders to describe the issue you mentioned? Or a comparison to the randomwalk2 from Arnold? Would love to see the differences.

[Bigguns]

You can see the effect I'm talking about in all the render that Lele did, they all look very bluish or greenish in the shadowed and valley region. I joined some render I did in corona with the volumetric sss there ( as I got better result then in vray, but still not good enought for skin, as it has the same problem as vray, greenish or blueish in the less scattered area... it's like if the reach of each rays it cut in the shadow region, like the sss has no effect sort of, it look's very odd.. but when the sss is strong across the whole object and the sss is scattered all over it look's GREAT , better then alsurface, but like I said, for skin we gonna have that much scattering, it would look like way too much and not realistic, so when going for a realistic strenght of sss, it then create dark area ( greenish or blueish depending the sss color you use.. ) If they can fix that problem, then it's the best sss hands down.

So in these image, hopefully vray forum did not changed the name of the pics, some are rendered in arnold with randomwalk2 ( wich is better then the one, the one had more greyish like the volumetric in sss in corona or vray, but still less though , the new one it's better, but not perfect ) it's better then the alsurface in this are of sss, like where the geometry is very thin, but it's less good in shadowed region I found.

Corona volumetric sss , when it's is strong and reach all part of the mesh , so there is no greyished area, it's AMAZING ! But unfortunately, not good for skin like I said and as you can see on gandalf face for exemple, in shadow region or valley's, it get's darker and greishor bluiesh..look,s really bad... so if the guy at vray can fix that problem, it would be the best sss out there.. but for now it's not...absolutely not..

Some of these tests are made with the alsurface for a comparaison ( some are noisy I know they were just quick test ) .

Another problem of the vlumetric sss, is trying to get the diffuse color working exactly like you'r texture.. good luck with that lol .. in arnold in work's well with the randomwalk2 and in vray with the alsurface it work's well too, you just put you'r diffuse into the diffuse slot and that's it. Also, it's veryyyy slow .

I would like if any prog from chaousgroup can comment on here and say if it's possible to fix these 2 issue with the volumetric so we can use it for even better skin then the alsurface.

So Jason, if you want to make skin, alsurface is still the best option by far in Vray.

[Bigguns]

Just noticied we can't post more then 5 pics per post :S I don't why but ok.. haha. so here's the other images ..

[vlado]

I'm having a deja vu from the time we discussed alSurface I'm afraid those issues are inherent to any subsurface scattering algorithm. The specific look might change slightly, but it's this effect that makes SSS look like SSS. You will find faults with random walk too.

Best regards,
Vlado

[^Lele^]

For the sake of completeness:

You can see the effect I'm talking about in all the render that Lele did, they all look very bluish or greenish in the shadowed and valley region.

They most certainly do, as one shader had a green SS color, and the other had a facing ratio gradient with a blue color, in one of the slots.
This, entirely to the side of the object of contention, the perceived quality of our RW implementation, on which i really haven't got the skills needed to comment on.

Attached the two shaders used across the different geometries (both were used on the murex shell too.).

[^Lele^]

Lastly, something else i can add to the discussion, and exploration of the shader properties, while not knowing the code: there is no inherent color bias, but rather it's entirely down to the setup chosen.
Saturated colors will produce very extreme shifts (especially with many bounces), so great care should be taken when applying maps and choosing hues.
Attached, a sample confirming there is no inherent "error" (or bluing/greening) in the shader, as the SS color has no saturation, all that comes through is the scattered color.
Notice how the scatter color too has low saturation, and yet it achieves very relevant presence in the final render.


Raising Saturation (typically past half way) will start introducing the debated shift: in my opinion the color IS way too saturated to represent capillaries, and so it stands to reason that other unwanted behaviour should happen.


Changing the SS color, however, will (correctly ?) virate the longest paths towards the complementary hue, with as much saturation as the color/map demand.


The same color in both Sub-Surface and Scatter will neutralise the effect by virtue of complementary colors adding up. It'll also make the shader very bland to look at.

[DaveKW]

Here are a couple of skin tests I did to compare the two. The scatterVolume had 20 bounces with a phase function of 0. Setting it above 0 just increases rendertimes and it was already double alSurface. 36 mins for alSurface and 1 hour 4 mins for scatter volume, bearing in mind that the scatterVolume had a refractive vrayMtl blended on top so that might be adding a lot to the render time.

The main differences I can tell is that there seems to be more contrast/definition in the shadows on the scatterVolume so the pores are darker, especially in the shadows and there also appears to be a darkening at glancing angles (this is most obvious on the bridge of the nose). The white artifacts on the scatterVolume render are in the reflection pass and I have no idea what's causing them although they are at the UV seams.


alSurface


scatterVolume



alSurface SSS render element


scatterVolume Refraction render element

[^Lele^]

Dave, the lips in particular look way better to me for VolScatter, but in general you're very right, the small detail is much better defined, and not as washed out as in ALSurface (which still looks pretty darn good, btw.).
Thanks for sharing a real world sample, very informative!

[Bigguns]

very good test Dave, So I guess you only swapped the SSS pass and kept the diffuse and reflection pass of the sss ? I never did it that way since I don't like to do passes haha.

We can see the back scattering in the ears is less good then alsurface to me, and the side skin pores becomes too visible, I think the alsurface too make them too visible ( the old sss2 was good for that ) since at glaring angles skin pores tend to disapears in real life. Also, that character has very smooth skin, with someone wrinkled , it would look dark and greyish in the wrinkles folds
with the volum scatter. The best result I got with the alsurface is 3 sss sheets in RGB color controlled via curves, that gives you control over the light path into the shader in each channel, work's slower of course then the single scatter, but yields to slightly more realistic result, especially in back scattering like ears or thin surface etc.

I joined an image, showing a sort of similar lighting and showing that the skin pores on the side becomes less apparent at glaring angles.

[DaveKW]

The beauty renders are 2 seperate renders copied straight from the frame buffer, there's no compositing or anything if that's what you mean. I only did the SSS and refraction passes to see the effects without any spec/reflection getting in the way.

[Bigguns]

ho interesting, how did you manage to get the diffuse looking the same color ?

[DaveKW]

I had to slightly lift it with an Output map as it did originally render darker. It wasn't too big a change in this instance as the phase function was at 0.