Page 1 of 9 123 ... LastLast
Results 1 to 15 of 124

Thread: Realtime Ultra realistic SKIN SHADER

  1. #1
    luxowaver silviotoledo's Avatar
    Join Date
    Jun 2003
    Location
    Brazil
    Posts
    1,831

    Realtime Ultra realistic SKIN SHADER

    Silvio Toledo
    Brazil

  2. #2
    Interesting approach, Silvio. Thanks. I have not tried that idea yet but I've been playing around with the screen-space SSS idea in the DP FNE:

    Lambertian diffuse shading (no GI) - 4.5 seconds



    multilayered-SSSS applied in screen-space with DP FNE in post-process - 5.2 seconds



    The real thing takes about 700% more time (38 seconds) - I haven't touched the default setup that comes with the model (simple skin material):



    the free model (and setup) courtesy of ten24

    Same screen-space SSS effect but with OrenNayar diffuse model:



    In my test, instead of using the final color render like in the documentation, I've used a custom-made illumination buffer. I've used also a different approach for the convolution, but tests looks somewhat promising. Guess I could improve the epidermis and sub-dermis scattering, perhaps.

    Would like to try also an implementation based in the real-time realistic skin translucency document you have linked, I suppose it would provide better results



    Gerardo

  3. #3
    Registered User jdomingo's Avatar
    Join Date
    Nov 2003
    Location
    perth
    Posts
    322
    just wondering why most cg character skin test or sss test are fare complexion? very very few in coloured as they called it. so what makes a perfect sss shader then?

  4. #4
    Not so newbie member lardbros's Avatar
    Join Date
    Apr 2003
    Location
    England
    Posts
    6,004
    I've seen someone here with some nice black skin shaders, Lee Perry-Smith did over at Infinite-Realities, think it was a Wesley Snipes like character, and was kindly given away on 3dWorld a few months or maybe a year or so ago, and then there's this one too:

    http://forums.newtek.com/showthread....real+character

    But don't think he uses Lightwave any longer unfortunately.
    Last edited by lardbros; 10-03-2011 at 06:13 AM.
    LairdSquared | 3D Design & Animation

    Desk Work:
    HP Z840, Dual Xeon E5-2690 v2, 32GB RAM, Quadro K5000 4GB
    Desk Home:
    HP Z620, Dual Xeon E5-2680, 80GB RAM, Geforce 1050 Ti 4GB

  5. #5
    Not so newbie member lardbros's Avatar
    Join Date
    Apr 2003
    Location
    England
    Posts
    6,004
    Quote Originally Posted by gerardstrada View Post
    Interesting approach, Silvio. Thanks. I have not tried that idea yet but I've been playing around with the screen-space SSS idea in the DP FNE:


    In my test, instead of using the final color render like in the documentation, I've used a custom-made illumination buffer. I've used also a different approach for the convolution, but tests looks somewhat promising. Guess I could improve the epidermis and sub-dermis scattering, perhaps.

    Would like to try also an implementation based in the real-time realistic skin translucency document you have linked, I suppose it would provide better results



    Gerardo
    Some lovely effects there Gerardo! How exactly are you achieving this? Very fast render times indeed, and looking very nice too! How complex are your FNE nodes are to get a look something like this?
    Last edited by lardbros; 10-03-2011 at 06:13 AM.
    LairdSquared | 3D Design & Animation

    Desk Work:
    HP Z840, Dual Xeon E5-2690 v2, 32GB RAM, Quadro K5000 4GB
    Desk Home:
    HP Z620, Dual Xeon E5-2680, 80GB RAM, Geforce 1050 Ti 4GB

  6. #6
    Quote Originally Posted by silviotoledo View Post
    those are beautiful renders i must say, interesting. waiting for Q4 2011 then :]
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  7. #7
    Gerardo, wonder, is screen-space SSS sort of like using the Z depth trick?
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  8. #8
    I proposed the use of the Z depth to achieve sss effects since 2004-2005 I think, but for controlling the absorbed color. In this case the depth buffer is used to control the blurriness, which achieves the scattering effect. It makes me remember somehow to my old trick for real-time blurred refractions, but in this method the blurriness is applied at different levels and driven by the depth buffer.

    I did pick that sample model because the lighting and shading setup is free and available for general domain and this provides a good reference for render time comparisons. I think an ideal model would be a similar free scene but with a female caucasian model. The reason for this, and I guess the reason why most of these tests use caucasian skin, is because the absorption of light in the epidermis can be more apparent in this type of skin. Forward scattering can be usually more notorious because lighter melanin (mostly pheomelanin) has a wider absorption than darker melanine (commonly eumelanin) - this highest absorption in the UV range makes caucasian skin more sensitive to UV rays and it's known that it's a good safeguard against solar radiation in african skin. Also if the hemoglobin in the dermis is high in the caucasian skin, backward scattering can be even more notorious, too. Both, melanine and hemoglobin, are key aspects in the skin color/tone and the reason why epidermis and dermis color are key parameters in the skin color of our models. Notice that although the epidermis in african skin usually contain mostly eumelanine (darker melanine), it could be also composed by greater quantities of lighter melanines, which makes this layer of the skin more translucent. There's also some cases where levels of hemoglobin (in caucasian and mostly asian dermis) are low, and backward scattering is not so apparent. The reason why a female model would be more suitable for these tests is because the structure of the female epidermis is commonly softer and more translucent, so scattering effects are more notorious, too.

    This is a test with an anim light:



    With the post sss effect:



    Would like to make more tests with other lighting setups to see the consistency and maybe improve a bit the backward scattering effect before thinking of sharing the node setup. If tests with other lighting conditions are promising, I'll post them here



    Gerardo

  9. #9
    ...forgot about that one, need to look into it again i think,... \:]
    http://forums.newtek.com/showpost.ph...5&postcount=18

    cool :]
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  10. #10
    I was actually referring to this one.

    I haven't had the time of improving the setup yet, but these are some side-by-side comparisons in different lighting conditions (lighting sets courtesy of Blochi's sIBL).



    Difference in render time is still less than a second. I've used Backdrop radiosity only but it seems the results of the sss effect in screen-space is consistent with direct and ambient lighting indistinctly. To try the consistency of the effect it would be interesting also test the lighting in other model.



    Gerardo

  11. #11
    looks great! :]
    i really wonder how far it can be taken...

    ok, hmm,... so am i right to say that you blur the depth map in post?
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  12. #12
    In the documents, the depth buffer is used to perform a depth-based blur removal so that the pixels far from the camera get narrower levels of blur. There are several implementations of the same principle but they commonly use a sum/average of about 6-7 one-dimensional gaussian convolutions, while in DP IFNE, I'm using DP Denoiser with several passes of bilateral blur weighted by the depth buffer, which provides good results so far. My setup is then not so fast but it's really simplified by DP Denoiser

    In my previous tests I was coloring (multiplying) the dermis and epidermis layers like in skin materials, but I've noticed this tends to change the overall color of the texture skin, and since it's supposed we are using a photograph-based texture, I suppressed the epidermis color in these lastest tests and I filtered each RGB channel differently for the epidermis and dermis layers with DP Curve.

    With one direct neutral light, this seems to work better than my initial setup. We can even notice the scattered red color in the shadowed areas.



    With neutral ambient color schemes (IBL) things seems work well too; however with this way of filtering the layers after the convolution, the skin looks not so multi-layered (a bit like rubber, I think) and too soft.



    Besides, with cold lighting, colors vary too much.



    A more 'correct' way would be to apply different weights for the blur function (convolution) of each channel, since red light scatters much farther than green and blue, we can suppose that the kernel of the red channel should be larger. This slows things down from 0.7 seconds to 1.2 seconds but results are more natural, I think.



    and the original skin color is preserved better as well (with direct or ambient lighting).



    ...and even with cold lighting.



    Still would like to improve the dermis layer.



    Gerardo
    Last edited by gerardstrada; 10-09-2011 at 09:10 PM.

  13. #13
    interesting, interesting,.. : ]

    ...another great thing about it is that it is flicker-free :]
     
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  14. #14
    luxowaver silviotoledo's Avatar
    Join Date
    Jun 2003
    Location
    Brazil
    Posts
    1,831
    This is amazing Geraldo!

    I need it!

    Newtek must include Mr. Dennis and you in the next Lightwave update .

    Step by step Tutorial pleaseeeeeeeeeeeeeee
    Silvio Toledo
    Brazil

  15. #15
    Adapting Artist jasonwestmas's Avatar
    Join Date
    Mar 2003
    Location
    North Carolina
    Posts
    11,399
    mmm yeah, that's crazy good skin there. Why is this not known until now I wonder. Is there some kind of underground development? What are the chances that we will be able to use something like this in our rendering package for animation? I mean the realtime skin, not the really nice 30s skin
    Last edited by jasonwestmas; 10-10-2011 at 03:38 PM.

Page 1 of 9 123 ... LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •