[eric nixon]

Problem on you renders

I understand your here to big up arnold? well thats a bit off topic dude...

Anyway Corona for c4d is available now, why would anyone use Arnold?



I used the same sss 'skinshader' technique here, plenty of contrast, mostly due to the lighting. Can Arnold render images like this?

Herve, I hope you dont mind me posting this.

[Aniki]

Not at all favoring Arnold here, just trying to push what Maxwell can do currently in terms of skin.

Herves Scene looked better I think. Back on topic as you compare apples and pears here..

This lacks texture detail:



Compared to rustebergs approach:



What I am expecting is something inbetween to be honest.

Below are my humble approaches, and I am not happy as well in terms of texture brilliance, contrast and detail perservation.

[eric nixon]

Actually thats my render, of Herve's model, (maxwell 2)

These new renders (must be maxwell right?) look very good, just the spec and disp is wrong. (did you use the trans map I posted?)

I dont know why you posted my initial attempt, rather than the tweaked version?

Basically I am trying to show that in order to get a physically correct render in a reasonable time, the substance needs to be more homogenous. As soon as you add diffuse bsdf's you get that heavy make-up look ( in this case horror make-up). I was also trying to show what can be achieved with a single bsdf.

To get a life-like quality the material must be driven primarily by transmission.

Please try lighting with hdri so that the realism can be assessed within some scene context.

Rusteberg never mentioned his rendertimes / sl needed, (and renders should be noise free by sl15-16 ideally) with no bright pixels after sl 3.. because this stuff is slow enough already.

Let me be clear, theres no way you can use maxwell to render skin for animations, until a new biased approach is devised.

[eric nixon]

This is the look I'd like to achieve, Image by Jorge Jimenez & co. from activision, I dont know which render engine, was shown alongside game footage, but cant be from a realtime engine. Her name is 'Lauren', maps by Paul Debevec & co.

[rusteberg]

Rusteberg never mentioned his rendertimes / sl needed,

don't recall...... it was somewhere around a gazillion hours. (Maybe somewhere between 2-4 on a dual big mama compooter - it didn't require overnight.....)

uhhhhhh....... the lack of detail & look of homogeneous structure comes from your sss because maxwell doesn't know any better.

with any and every sss mat i first create the base:

0,0,0 rgb in both reflectance slots
0 roughness
1.001 nd

then work on attenuation value (number value) and color to get the amount of light i want transmitting the surface
then fill that substance with a an amount of light diffusing through it (Scatter Coeff), a color in which to exit each scattering particle, and a factor of how dense each particle is (assymmetry)


soooooo........ i was getting frustrated that i couldn't get the sss depth i was looking for in solid sss objects as shown here where the steak looks a bit uniform like some of those head tests



so i added more color to it with an additive layer (and as always controlled the spec with its own layer on top of stack)
here you can see the steak appears to have more "depth"



and it seemed to work well for what i wanted it to do as a universal base skin shader among various models including the head...... minus that weird geo thingy..... ( i find that 2.5 - 5.0 is attenuation value range depending...... although the brightness and contrast of map can make that value relative - and as sensitive as the scattering assymettry which is why i focus on that value first and then "fill" it with substance.... fyi)





tadaaaaaa!




( . ( . ) boobies

[Aniki]

Thanks Rusteberg! Always good to get some decent second approach on things.
On the Windowlicker project I also dialedin more saturation and contrast to counter the washout.

I still guess the current solution would involve to mixx several sss bases with different attenuations/scattering/direction variances to achieve that distinctive skin scattering look.

[eric nixon]

Some progress towards usable non-homogenous sss materials ; seems like weightmapping is allowed (still slow but no 'dusty' perma-noise) if certain parameters are shared between bsdf's.
All renders of the same mxm. I added a black skull object which increased the benchmark, and the larger render has disp turned on.

[eric nixon]

Tweaked version, sl16, about 2hr's with displacement on. (changed the 'rules' a little - different roughness maps..., now using Rusteberg's low roughness idea with low spec in the thicker bsdf)

edit: see SL19 image below.

[bitarafali]

Cooool
Really good eric and good job!!

What map you use for weight?

[eric nixon]

Just painted it in Photoshop. Its just a guess.

SL19

[Aniki]

Now we are talking. This is your best result yet. I still miss a little saturation/coloring in the shadowy areas-

[Gokhan]

Dear Eric Nixon,absolutely great renders my friend,I loved it,
Someday I wonder if I can get that are rendered?
I'd love to get such high quality rendering .

[eric nixon]

Thanks Gokhan, bear in mind these maps we're using have a lot of baked light polluting the map. That will help the appearance of realism from certain lighting angles, less real in other light. Another model with good geo, like this one, but with better neutral maps, could have a more even realism in all lighting, but not look as real as this one does in this view...., and also the maps are very important, and will often need adjustments per model.

Repost:

[choo-chee]

this looks too good to be rendered. that's a real dud right ?

[bograt]

Here are some old tests of mine:

Maxwell: ~ 18hrs.


Vray Fast SSS ~10mins. FYI, Color profile was not compatible with web so looks very desaturated


Maxwell's sss will wipe the floor with vray's in terms of accuracy but it is very slow for such shallow scattering. Corona now has scattering but it is still being developed... like Vray it is very fast but not so physically accurate as Maxwell.