Screen-space human skin implemented on V-Ray for 3ds Max that reduce render time 90%

[vlado]

Almost forgot; about the harsh shadows that you are getting when you set the phase function to 0 - these are due to the single scattering approximation; if that bothers you, just set it to "None" (or perhaps raytraced).

Best regards,
Vlado

[Bigguns]

Hi Vlado,

thanks for you answer.

When I mean get inspired by real life references, I mean, for the look only. You know better then me the maths behind the shader,etc. I just try to help with what I see can be improved, it's you who knows how to make the changes in the shader to reflect that. Brute force sss seems good but maybe it gonna be slow(slower)?

I tried for the ears to use a map for scatter radius... ok the ears was looking not too bad when backlighted.. but in normal lighting, they seems to be in wax! because the radius was too big on them... Is that complicated to make the backscatter control like the one in Mentalray ?
The only way I got a not too bad effect while keeping it correct in other lighting, was to blend it with a Vraymat and using refraction with fog parameters.. but it only work with direct light, with indirect light the effect is not as good as it should be.

I tried what you said here : about the harsh shadows that you are getting when you set the phase function to 0 - these are due to the single scattering approximation; if that bothers you, just set it to "None" (or perhaps raytraced).

And then it gave me the exact same result has if I was not putting the phase function to 0 etc. I got the same result as the phase function at 0.9 and the approximation to single... with the greenish bug... so is there something that is breaked in the sahder? Try it yourself you will see, it gave the exact same result :S There is no workaround to do to not get the greenish bug... the only way to alter it, is to use a color scatter close to your sss color and use a map for the radius, where you have thin area like wrinkles and eyelids,etc and map those area a little bit darker...

The file is easy to find on the net ( but this model head is not the best to test a skin shader, too soft surface, better to have a realistic old man head with wrinkles..) , but you won't have my set up though... I custom made glossy map and reflection map.. but anyway, don't really need them, it's more the SSS were talking about here but sure adding control of the `soft` fresnel reflection would be awesome instead of having to blend it.. and opacity map pleaseee would take you 5 min to incorporate it and would be awesome for all of us

It's really important to upgrade constantly as much as you can the skin shader, it's one of the big things that made peoples switch to Vray like me haha. I was using Mr ONLY because he had a skin shader... so when Vray got a better one.. the switch was instantly lol like a lot of peoples are doing since 2 years..(and I knows studios that are testing it right now to see if the skin shader suit their needs ) since they saw that now we can make good looking skin in Vray.. it's a VERY important factor. So if you can get the best skin shader out there.. the more customer and big studio you will attract.. now you got blur, since they saw that Vray had a nice skin shader but more will I believe Vray is really cool but don't do like Mentalray and taking 5 years to upgrad shaders haha (they did a new sss2 with the latest Mental ray.. it's better then the original..but it still have the same flaws.. looking plastic.. with some red in the shadows etc.. the bump suck, look like it's not scattered by the sss,etc..)

The coolest thing about the SSS2 in Vray is the way it handle the front scatter (when you don't have the greenish effect of course) we really feel the depth of the skin, and that is really cool! I think in that departement, Vray rules Arnold and all of them. but the other little bug make it more bad in some circumtansce, like low back scatter effect,greenish effect when scatter is a bit strong, reddish bounce of light,etc. So if you can correct these, I think you will have the best skin shader out there

P.S edit : An other thing that is working really well in the shader, is the specular and the way the bump react with the specular and the sss, the bump is affected by the scatter wich make it well integrated togheter and flows naturally, like in true skin and.. it's so simple to get the diffuse right not like mr that you need 3 maps to get about the resutlt of your diffuse (while never really getting it exactly..) that in Vray sss2 work perfectly! the diffuse color are really well preserved and not diluted in the sss and also then displacement is not diluted.. in mr all the details goes out... so when upgrading the sahder, make sure these things are not broken along the way the good's must keep as good as they are, just upgrading what need upgrad would be perfect

[Bigguns]

Hi Vlado,

me again,
I still think you should really look at this PDF closer : http://delivery.acm.org/10.1145/1970...33106054c48f08

It has solve the greenish problem of Vray and the boucing redish bug of sss in Vray sss2. Vray use a dipole I may guess because it look quite ressembling to the reference image on the left in the exemple... and on the right is the new Quantized-Diffusion method, wich is more accurate, you can tell already by seeing the lips, they look less waxy and the shadow are much better, I'm pretty sure this model work perfectly with wrinkles too ( no red bug in thin wrinkles..)

If I was you vlado, I would let the sss2 as is ( just adding opacity map ) since it's more of a multi purpose SSS shader and create a NEW TRUE skin shader that is dedicated really for doing skin and I would based it on this PDF, on this Quantized-Diffusion method. That would be simpler for you I guess and better for us I cannot force you to do that but I speak in the name's of many, let's say that like this haha.

It's interesting to read in the paper that they said this : The new diffusion solution accurately decouples single and multiple scattering. We then derive a novel, analytic, extended-source solution to the multilayer searchlight problem by quantizing the diffusion Green’s function.

I don't know if there talking about the same green as I refer that I hate in sss2.. but sound's like it is... so with this new shader you get rid of it.. as you can cleary see here in the rendering of this alien : http://www.eugenedeon.com/octoguyLarge.jpg from this webpage : http://www.eugenedeon.com/qd.html you can cleary see that, even with a BIG RADIUS, the greenish effect is not there, instead it's yellowish, as I told you earlier.. and that is way much more accurate.


This shader combined to the superb GI quality of Vray will be the killer combo for realistic rendering of skin, believe me It will put Arnold and the other's to the stone age in rendering realistic character haha In fact I think it will be so realistic that it will be hard to improve it again more. And lot's of studios will use Vray for it's ability to not just do archi viz (wich everybody know Vray rock at it), but hyper realistic rendering of characters,creatures, etc( don't get me wrong here,sss2 is pretty cool, but it can be cooler and need to make it the top skin shader) , it's a very big and important aspec for the big studios ( so you won't just have mainly Arch viz client, but also VFX studios,etc ), even more for smaller one too.. since they don't all have the budget to make their own shader.

Some other references on it : http://dl.acm.org/citation.cfm?id=1964951&bnc=1


You the man Vlado, it's up to you to do a dedicated skin shader using this Quantized-Diffusion method, but in the end if you do that, your customers will be very happy.. and so much more to come I try to help as I can and if you need me to test and give my input about a new skin shader (or upgraded sss2 ) let me know and I'll be glad to give my artistic point of view on it while testing it. I have some models really realistic so it's nice for testing the skin shader too.

All the best

[vlado]

I've read it in detail, yes. However, I would really rather fix the sss2. Since you seem perfectly happy with the Arnold shader, I'm sure we can make sss2 work for you as well However if you have trouble controlling this relatively simple shader, then I can only imagine the trouble you'll have to control the quantized diffusion shader unless you dedicate a lot of time on set up trials.

Best regards,
Vlado

[Bigguns]

Hi vlado,

I have no problem to set up the sss2, it's pretty simple in fact ^^

What I mean, is we can't get around some bugs or limitation of the shader itself, like the green effect etc.. no matter what you do.. you are limited to the shader..

Before using vray sss2, I was using mentalray one.. wich is way more complicated to use, I had generally at least 6 maps to 8... in Vray I have 4 , one diffuse, bump,(and displacement) spec and glossiness map. No need for more, maybe a scatter radius map to control some more it ..

Arnold skin shader is not perfect... as I had see, it do not scatter the bump with the sss,wich give it a harsh look.. a la Mr(for the bump) .. :S.. Vray does better here... as I said, in Arnold what is cool is the back scattering working better and no green effect in the front scatter.AND you can build your own,..that is amazing to me.. so you can customise it for your needs.. (but if you do the skin shader perfect, we don't have to customise it haha) , other then that, I prefer vray sss2 and ... Arnold is way much faster to render skin.. but not the rest lol (like GI haha) Vray skin is not so slow, but not so fast.. it became realllyy slow when I use scatter indirect light with GI... now that is slow.. can you make this faster? I try to avoid using GI with sss2 cause of that point, try to fake the gi with direct illumination.

Don't bother with it, I can control very complicated shader, skin or not so if you have a Quantized-Diffusion shader ready for me to test it and show you what I got compared to the sss2 , let me know

[vlado]

What I mean, is we can't get around some bugs or limitation of the shader itself, like the green effect etc.. no matter what you do.. you are limited to the shader..

Well, this might be a bug, in fact. There is something odd, but I can't put my finger on it yet - for the moment, I'm re-checking all the maths...

Best regards,
Vlado

[Bigguns]

Good news Vlado, really glad to hear your working on it

I believe in you haha

What I like about chaosgroup, is that your listening to your customers, and this, in the end, always pay back

So again, when you have something changed in the sahder, if you need me to test it and give my feedback or recommandation on it, I'm open anytime

thank's!

[Morbid Angel]

I've read it in detail, yes. However, I would really rather fix the sss2. Since you seem perfectly happy with the Arnold shader, I'm sure we can make sss2 work for you as well However if you have trouble controlling this relatively simple shader, then I can only imagine the trouble you'll have to control the quantized diffusion shader unless you dedicate a lot of time on set up trials.

Best regards,
Vlado

Arnold's skin is nice, but not without its own problems.

Vlado, SkinSSS3!

[Bigguns]

How things are going Vlado about the skin shader? Any infos to share about it?

[vlado]

There are some problems that I found, and in the meantime I also coded a brute force SSS shader to compare the results; I hope I'll be able to post soon.

Best regards,
Vlado

[Bigguns]

Fantastic news Vlado glad to hear that!!

And for this new release, just as a reminder, don't forget to add :
an opacity slot
Back scatter control (if possible)
Full reflection control

Can't wait to try it out!! I'm on a new character and it will be very usefull I'm sure

All the best

BTW, I normally go into the vray properties of the object wich has the SSS2 on it, let's say the face, and put the generate GI to 0, it help a lot to have less bouncing light effect between close surface, it's not perfect but better.. But can't wait to try your brute force one :P

[anchovy]

Any chance the brute force version could be used withthe PPT or BDPT?

Thanks

-Michael

[vlado]

Any chance the brute force version could be used withthe PPT or BDPT?

Yes, it already works fine with the regular path tracing; the bidirectional path tracer is a bit more work, but we'll get there too.

Best regards,
Vlado

[anchovy]

Yes, it already works fine with the regular path tracing; the bidirectional path tracer is a bit more work, but we'll get there too.

Best regards,
Vlado

Fantastic!

[vlado]

Hehe, don't get too excited, it's terribly slow since it does hundreds of brute force light bounces inside the material I still have some work to optimize it.

Best regards,
Vlado