Pixelvore, yea using screen is better, even great mental ray skin shader use screen compositing (i think so but i’m not sure) if u want make good specular try using phong, i know that phong isn good for that, but by mixing some phong shaders (two or more) u obtain nice results this little trick is also used in mental ray skin shader:), also i dont see in your node tree any setup for fresnel;) one more thing dont set bump on epidermal and subdermal layer bump SHOULD be only on unscatter diffuse, good balance between unscatter diffuse and epidermal layer is key to nice looking skin.
Soon i try make some tests.
Thanks Maqs for your information, I’m really glad to see that with the help of other people, I can improve my shader, and in the same time increase my knowledge 
About specular shaders : I already use 2 specular terms, one of the speculars layer is set on the epidermis (soft looking speculars in order to lighten the bumps) and brighter speculars on another shader. About the fresnel effect : I will do a fresnel setup if really needed, for example if I need “velvet-like” skin (or maybe it could be done with some nice looking fur simulation, with almost microscopic strands - but it would imply huge amounts of strands). Moreover, the “fresnel” shader or even “minnaert” produces interesting fresnel effect. However, I really want to keep the “oren nayar” shader, since it takes into account the microscopic bumps of the surface, changing slightly the appearance of the skin. This way, I obtain quite interesting effects on the surface, and in the same time, under the surface with sss. For the speculars fresnel, I saw that there was already a kind of fresnel effect on the “cook-torrance” shader, just take a look at this image : exactly the same specularity, only the direction of the incoming light is changing. With the Phong shader, there is no fresnel simulated, no matter the light comes from, the speculars will appear normal, whereas with the cook-torrance speculars, the “edges-speculars” appear brighter :
About the bump texture, and where it should be applied… I also noticed that sss, even when used very slightly, tends to blur the bump a lot, and that’s why I use a very light sss effect on the epidermis, and if you only render the epidermis, you’ll notice that it’s almost an unscattered diffuse. I’ll try what you say, I mean, using unscattered diffuse + epidermis in the same time (in my mind epidermis scatters the rays so slightly that it’s almost an unscattered diffuse… When browsing the indigo forum, I saw that scientific measurements tend to show that epidermis is really thin : 0.35 mm ! so it cannot scatter light a lot.)
Yes ofourse epidermal/epidermis should be only slighty scattered, but withoute unscattered diffuse layer, light flow isnt simulate properly, becouse all layers (epidermal and subdermal) absorb and bounce light, but there also should be layer that only bounce light.
I make quick graph in amazing painting sofware, Ms paint.
http://images25.fotosik.pl/193/3b1159f3b3ec476dmed.jpg
try search some graph from misss_fast_skin shader, i think this is what u need now.
edit.
oh About spec i mean something like this (phong)
http://images32.fotosik.pl/214/0d2981a0218948a5.jpg
Thank you, I had seen some drawings like this and since epidermis doesn’t look good by now, I will follow your instructions about adding another layer on top of the epidermis, which will be textured.
About speculars : the use of 2 specular shaders is already implemented in my shader, and yet, imo, the speculars still don’t look convincing 
Hey Pixelvore, here you can read up on BRDF too:
www.cs.kuleuven.ac.be/~phil/GI/TotalCompendium.pdf
Wow, 68 pages
…
I’m gonna read it carefully ! (not kidding 
)
This is not plausible. If you are looking for a realistic reflection model, then that is not it. It does not makes sense that the object and the reflected image of the object be the same intensity if the reflecting object is anything darker than pure white.
You always need to take into account the law of energy conservation. A surface receiving light cannot absorb and reflect more light than it receive. If your object surface is 10% gray, that means that it absorb 90% of the light it receives so it cannot also reflect 100% of that light. It can only reflect 10% on that light. So all of the square planes would be reflected on the cylinder to only 10% of their actual intensity.
This is also a common mistake made by CG artists with specular shaders too. The balance between the diffuse reflection and the specular reflection is oftentime way more than it would bein reality. Normally, the diffuse and the specular component addd together should not be higher than 100%. So if your diffuse color was 10% gray, then the specular component should not be higher than 90%.
The same principle would apply to the skin shader in the sense that the total of all the layer reflectances should never be higher than 100% reflectance.
I would need a strong proof of that to believe it. You may be getting good result here in this test situation but as I mentioned before, Screen operator assumes a few things about the RGB values and I predict that those assumptions will eventually get you into troubles in different lighting condtions such as HDRI IBL and/or when trying to apply tome mapping operators. But this is yet another assumption from me though. I would like to be proven wrong on that. One thing is sure, I will not design complex maps to use on a head or body model that uses a shader model with such potential issues. I prefer base my design on a shader model that is more robust ant that can theoretically hold on under more extreme lighting conditions and tone corrections. That is just me though.
Indeed, I noticed that when using a very strong lighting, when both dermis and epidermis appear white, the screen node gives oversaturated blue… But since I do not intend to setup any lighting like this (I mean, such an ugly lighting setup :D), I think screen is the best. But I have to admit that it seems strange that Mental Ray does use a screen operator to combine epidermis and dermis… However, I’m going on with my tests with the Mix node, and they are still flat, plastic looking, instead of skin
edit
Take a look at this… Incredibly convincing, innit ? (a render with the screen node, with a far too strong lighting)
What about tweaking the subdermal shader parameters to get the proper skin look you are looking for then?
That’s what I’m trying to do by now, I’ll post some renders when I have something correct looking.
Some tests.
http://images33.fotosik.pl/215/30969f667638107d.jpg
http://images26.fotosik.pl/194/200a68d7b1a5737a.jpg
ehh… woah!!
maqs, is that using the same material made by Pixelvore ?
really great stuff.
No, i use my own node tree, but its similar to Pixelvore shader, i try made more tests with better models and textures later.
I would need a strong proof of that to believe it…
as i say im not sure about this, but if u chose fast skin shader, there is option screen composit turned on, but u can turn it off also, if u search web u can find a graph that show how mental ray shader is build and there also is write something like this.
Epidermal (add) Subdermal
add= soft screen composit
sorry but i realy cant find that graph, here is node from blender two mental ray.
http://images24.fotosik.pl/195/57f4e2c143196a7f.jpg
screen_composit, when turned on, chooses Screen compositing. When simply adding together the contribution of many layers of light, it is very easy to quickly blow out and overexpose into white, but the human eye is inherently nonlinear and perceives intensities in a different way. This option allows use of what is known in many compositing applications as a screen transfer mode between the layers, which yields a softer, more pleasing result. If it is turned off, normal addition is used. If rendering in a high dynamic range, and a proper tone-operator is applied in the final output stage already compressing the final luminance, use off. If not, turning this option on yields a more pleasing result.
So we are both right :). This is an ON/OFF option. And like I said, Screen mode should not be used when using strong lighting schemes (HDRI) and/or tone mapping. But it gives better results.
Your skin nodes does a hell of a good job. And your models are really fine too.
Thank u ypoissant, i think i done with 3 layer sss, now i need work on other things especialy specular maybe i find some papers aboute this, i want also compare my shader with mental ray fast skin but unfortunalety i cant corect join shader components in mental ray 2 blender build 
Edit.
here is node tree i use for tests renders, fresnel is turned off, and i dont use back scatter map and overall becouse that was only quick tests.
Can’t someone just write a Pynode script to aid in greatly reducing that node tangle?Download SVN and the Pynode is under the dynamic catagory.
Just Do It, CD.
“Can’t someone just write a Pynode script to aid in greatly reducing that node tangle?Download SVN and the Pynode is under the dynamic catagory.”
Its to early;) shader isnt finished but when I/WE finished it some dev could upgrade standard blender’s sss base on my/our node tree or add new option “multi layer sss” or “Fast skin shader” something like that:) but as i say its to early to do it now. I wait for Pixelvore move now maybe he develop any tricks for realistic specular:)