3Delight Surface and Lighting Thread

13468952

Comments

  • wowiewowie Posts: 1,987
    edited December 1969

    wowie your getting some neat things to happen,

    Mustakettu85 thanks for your interest give me a few more days, I get busy durning the week, but on the weekend, hopefully sometime Sat. I can get something on sharecg. Since my work is piliminary, I think after youve downloaded it, I'll remove it from sharecg

    wowie if you use pwsurface your welcome to have a look too, I think were the only ones left in the thread?

    Thanks. There's some newer tweaks, both to the material and lighting. I haven't had the chance to render and post them here yet.

    Unfortunately, I don't have PWSurface. But I think some of my preset can be ported/translated into PWSurface.

  • Mustakettu85Mustakettu85 Posts: 2,917
    edited December 1969


    Mustakettu85 thanks for your interest give me a few more days, I get busy durning the week, but on the weekend, hopefully sometime Sat. I can get something on sharecg. Since my work is piliminary, I think after youve downloaded it, I'll remove it from sharecg


    That's fine with me =) Waiting with interest!



    Human skin when viewed under low light shouldn't be that dark.

    Below you'll see my setup under similar conditions as your third render. They are not using area lights and I've turned down UE2 strength to 25%. For the second one (the top one), I dialled UE2 strength down to 5% and upped the spotlight to compensate. I think this one is closer to yours.

    Although the ears do have shadows, it's nowhere near as dark as yours. You can also see the shadows on the left side of the face are not as dark as well.

    What's the shader you're using and what are the SSS settings? It would be interesting to see your settings under the same exact light I used (if you're willing to try the IDL route, please get the whole scene sans Genesis here: http://www.sharecg.com/v/69379/view/21/DAZ-Studio/Mk85-Basic-IDL-scene-for-closeup-portraits ; if you only want the AO, it's here: http://www.sharecg.com/v/69281/view/21/DAZ-Studio/Mk85-Basic-Lightset-for-DS4.6+ - just set the shadow bias on the "shader" light to 0.1). And if you're using Genesis, let it be the same pose for consistency... it's one of the "basic female" ones, the Walking 01 one.

    Or alternatively, you could post your lights somewhere...

    Your images are so heavy on translucency!.. Similar to the "wax figure" look the Poser folks prefer these days. "Hyper real". If that's artistic intent, then it's perfectly fine, it's a popular style these days because it has that undeniable "prettiness". It's just that the only real life example I can remember that would look somewhat like that is Laura Bertram playing Golden Trance... with a thick layer of glittery-n-scattery makeup all over her body: http://trance-gemini.com/wordpress/wp-content/uploads/2011/11/trance1001-copy-757x1080.jpg
    Do you have any other real life examples of such... "delicate" flesh? (no, really, when I look at those renders of yours, the words that immediately pop into my mind are "daintiness", "fragility", "a delicate flower of a woman", that kind of associations; I'd use your settings for rendering faeries or Tolkien-type Elves - beings who are more spirit than body).

    What I am generally aiming at, though, in my work, is not "prettiness" (I confess ignorance for most things "pretty", actually, this is why I'm asking you for your real life references) but "drama", for lack of a better word. Think chiaroscuro in painting: http://en.wikipedia.org/wiki/Chiaroscuro - as for real life examples of the style I'm generally trying to imitate, there's that Canadian movie, it's probably not that well-known, but it has a handful of my favourite actors in it: Tactical Force. The poster ( http://www.imdb.com/media/rm1498657280/tt1458915?ref_=tt_ov_i ) only hints at how the movie is actually filmed, which is very high-contrast and probably heavily post-processed - exactly what I've always wanted to see on screen. There are many similarly lit interior shots in Gene Roddenberry's Andromeda and Muhteşem Yüzyıl, too (these are the shows that truly grabbed my attention with the "graphic" way lighting is handled in them). It's probably because whatever little formal artistic training I've had was mostly for graphite pencils, and you know it's a medium that thrives on contrast and "drama".

    //and IRL, people around me aren't actually pretty faeries but waaay more "rugged"-looking =D//

  • wowiewowie Posts: 1,987
    edited December 1969

    Finally had the time to do some quick renders.

    It's just plain HSS for the shaders, UE2, plus UberAreaLights and one standard default distant light. I mainly work in DS3. Have DS4, but always seem to crash when I tried to render (randomly).

    The materials and lights still need some tweaks. As you noted, the transluency is pretty heavy and can be too much. That's mainly to setup the upper limits for the lights.

    The materials and lights work best together. The materials will likely show up overexposed/underexposed with most light sets available (free and commercial). The main purpose for the materials and lights was to avoid having that Poser/DS look, and strike a good balance between quality and render times.

    Render_5.jpg
    800 x 1040 - 233K
    Render_3.jpg
    800 x 1040 - 237K
  • wowiewowie Posts: 1,987
    edited May 2013

    Another render, using pretty much the same light. The difference is that I turned up diffuse strength and lower diffuse color by about 10% (strength up, color down) for the second render.

    Render_10.jpg
    800 x 1040 - 226K
    Render_8.jpg
    800 x 1040 - 222K
    Post edited by wowie on
  • Mustakettu85Mustakettu85 Posts: 2,917
    edited December 1969

    wowie said:
    Finally had the time to do some quick renders.

    It's just plain HSS for the shaders, UE2, plus UberAreaLights and one standard default distant light. I mainly work in DS3. Have DS4, but always seem to crash when I tried to render (randomly).

    The materials and lights still need some tweaks. As you noted, the transluency is pretty heavy and can be too much. That's mainly to setup the upper limits for the lights.

    The materials and lights work best together. The materials will likely show up overexposed/underexposed with most light sets available (free and commercial). The main purpose for the materials and lights was to avoid having that Poser/DS look, and strike a good balance between quality and render times.

    Super pretty! =) HSS does seem to handle SSS differently from UberSurface, I wonder why... were there any undocumented code rewrites under the hood? Ah, Omnifreaker only knows...

    And yeah, DS4 tends to crash when it wants more memory (which, on systems with 2GB RAM or less, is almost all the time when the scene is any more complex than a figure and UE set to AO...). What I do then is render to RIB files which are then easily rendered through the standalone 3Delight.

    The lights vs complex materials is a giant problem actually, it's true. You'd think that it were avoidable if we tried to stick to physics-based stuff, but there are way too many variables in unbiased rendering to really ensure universal presets.

    At home, I was testing AoA's shader further and loaded my "go-to" lightset based on my own HDR map that I routinely use for US2 renders. Pictures speak louder than words, so here's a giant sheet of various effects for comparison. In short, you gotta optimise your lights when you switch shaders, because man, do they render differently!

    SSS_tests2000.jpg
    2000 x 1646 - 495K
  • wowiewowie Posts: 1,987
    edited May 2013


    Super pretty! =) HSS does seem to handle SSS differently from UberSurface, I wonder why... were there any undocumented code rewrites under the hood? Ah, Omnifreaker only knows...

    I think the SSS brick is the same for all three (HSS, US/US2 and AoA Subsurface). The controls are different though. I like HSS more because there's not actually that much to work with. The addition of scatter and absorption color in US2 makes it work differently from HSS, but it's still tied to diffuse strength (I think).


    And yeah, DS4 tends to crash when it wants more memory (which, on systems with 2GB RAM or less, is almost all the time when the scene is any more complex than a figure and UE set to AO...). What I do then is render to RIB files which are then easily rendered through the standalone 3Delight.

    Yeah, I really need to upgrade to a proper 64 bit machine. I think DS4 bumps into the limits of 2 GB (the amount of RAM I use) way too much compared to DS3, particularly when rendering. Just waiting for Haswell to be launched next month.


    The lights vs complex materials is a giant problem actually, it's true. You'd think that it were avoidable if we tried to stick to physics-based stuff, but there are way too many variables in unbiased rendering to really ensure universal presets.

    I think you mean biased rendering. I personally don't think biased or unbiased makes much of a difference. I've seen just as many bad renders with unbiased renderers and good ones with biased renderers. Ultimately, it's up to the artist to tell the renderer what he/she wants. They need to understand how to coax the best results of the tools they're using.


    At home, I was testing AoA's shader further and loaded my "go-to" lightset based on my own HDR map that I routinely use for US2 renders. Pictures speak louder than words, so here's a giant sheet of various effects for comparison. In short, you gotta optimise your lights when you switch shaders, because man, do they render differently!

    The optimized lights/materials looks better.

    A point you might want to consider when making materials is not having UE affect scatter (too) much. Put it more simply, you're seeing UE rather than SSS. I made this mistake before. That's why I now always test with two lighting setups - one with soft diffuse light (much like photo studios) and another with harsh lights with very low UE intensity). Once you figured it out, you have a skin/materials setup that's pretty flexible regardless of lighting.

    Another is having both light and dark skin characters in the same space. Your materials setup should be flexible enough that even in shadowy areas, dark skin is still generally visible (or barely at least) With preSSS methods, you generally have to lower SSS strength a bit. I'm thinking with the postSSS, you don't need to do that.

    Edit. After rendering the shots below, I think I can have the same SSS settings and just rely on the diffuse color/strength combo to control the amount of SSS applied.

    Render_9.jpg
    800 x 1040 - 299K
    Render_8_(2).jpg
    800 x 1040 - 279K
    Post edited by wowie on
  • Kevin SandersonKevin Sanderson Posts: 1,637
    edited May 2013


    At home, I was testing AoA's shader further and loaded my "go-to" lightset based on my own HDR map that I routinely use for US2 renders. Pictures speak louder than words, so here's a giant sheet of various effects for comparison. In short, you gotta optimise your lights when you switch shaders, because man, do they render differently!

    Good for you pointing that out. I see so many not working on their lighting and expecting instant great results with the shader. The other thing I keep coming across, except for a couple of people, is the fact that folks aren't referencing real photos to see how far off they are from real skin. Not a problem if you don't want real skin, but for those who do, it's very important.

    Post edited by Kevin Sanderson on
  • RAMWolffRAMWolff Posts: 9,679
    edited December 1969

    Cool thread. Subscribing...

  • Type 0 NegativeType 0 Negative Posts: 320
    edited May 2013

    Thank you Sincerely for all your work!
    Why do you call them "crazy"?

    I'm using them now, but I did a few tests and turned the Occlusion Strength down to 10.0% with Occlusion Samples at 24
    may just be a personal taste thing?

    Mustakettu85, do you have pwSurface2 installed?
    If so and your interested in what I'm doing here, I'll get this skin material setting to you.
    You might be able to tweak it up better.
    I'm using some strength images I painted, and I've reapplied the diffuse textures to other areas of the surfaces, added them to Translucense Color and to Subsurface Color. I have not figured out how to remove the strength images I painted to get similar results.

    You're welcome! Why "crazy" settings... well, because these settings add longer render times and the quality improvements are rather subtle =) This is where I use the word "crazy" =)

    Yes, I have pwSurface2, but I don't use it often... and judging by your renders, especially how noticeable the blue on the Translucence channel is, it works differently in some aspects than the omnifreaker shaders... So yes, I'm interested!

    Okay I got it ready. As soon as you let me know you have the download I think I will take it back down, because it is a work in progress.
    If anyone else wants it feel free to get it now.
    http://www.sharecg.com/v/69410/browse/21/DAZ-Studio/Lana-preliminary-skin-and-lights-for-pwSurface2
    hope it works and can be tweaked better by you
    good luck and have fun

    Post edited by Type 0 Negative on
  • Kevin SandersonKevin Sanderson Posts: 1,637
    edited December 1969

    Thank you Sincerely for all your work!
    Why do you call them "crazy"?

    I'm using them now, but I did a few tests and turned the Occlusion Strength down to 10.0% with Occlusion Samples at 24
    may just be a personal taste thing?

    Mustakettu85, do you have pwSurface2 installed?
    If so and your interested in what I'm doing here, I'll get this skin material setting to you.
    You might be able to tweak it up better.
    I'm using some strength images I painted, and I've reapplied the diffuse textures to other areas of the surfaces, added them to Translucense Color and to Subsurface Color. I have not figured out how to remove the strength images I painted to get similar results.

    You're welcome! Why "crazy" settings... well, because these settings add longer render times and the quality improvements are rather subtle =) This is where I use the word "crazy" =)

    Yes, I have pwSurface2, but I don't use it often... and judging by your renders, especially how noticeable the blue on the Translucence channel is, it works differently in some aspects than the omnifreaker shaders... So yes, I'm interested!

    Okay I got it ready. As soon as you let me know you have the download I think I will take it back down, because it is a work in progress.
    If anyone else wants it feel free to get it now.
    http://www.sharecg.com/v/69410/browse/21/DAZ-Studio/Lana-preliminary-skin-and-lights-for-pwSurface2
    hope it works and can be tweaked better by you
    good luck and have fun

    It looks nice! Thank you!

  • SloshSlosh Posts: 2,388
    edited December 1969

    Where is wancow? Haven't seen hide nor hair of him in ages. I hope all is well, if you are out there reading this, buddy.

  • Mustakettu85Mustakettu85 Posts: 2,917
    edited December 1969


    Okay I got it ready. As soon as you let me know you have the download I think I will take it back down, because it is a work in progress.
    If anyone else wants it feel free to get it now.
    http://www.sharecg.com/v/69410/browse/21/DAZ-Studio/Lana-preliminary-skin-and-lights-for-pwSurface2
    hope it works and can be tweaked better by you
    good luck and have fun

    Thanks a lot, I got it! There's going to be a lot of stuff going on at work for the next few days, but I hope to have some quality time to devote to pwSurface towards the end of the week.

    wowie said:
    I think the SSS brick is the same for all three (HSS, US/US2 and AoA Subsurface). The controls are different though. I like HSS more because there's not actually that much to work with. The addition of scatter and absorption color in US2 makes it work differently from HSS, but it's still tied to diffuse strength (I think).

    There are some differences in the way SSS treats diffuse (US2 can have diffuse off, but with some strength dialed in regardless, the earlier shaders want diffuse to be on), but as we don't have the source code, we can't say if it goes any deeper than that...


    wowie said:

    Yeah, I really need to upgrade to a proper 64 bit machine. I think DS4 bumps into the limits of 2 GB (the amount of RAM I use) way too much compared to DS3, particularly when rendering.

    Yup, it's rather capricious. Smoothing is another memory-hungry feature. Still, even with a measly 2 GB, a 64 bit system can render more complex stuff (like RiCurves) than a 32 bit one, without crashing.

    wowie said:

    I think you mean biased rendering.

    Yes, sure. Sorry, that's what you get during a heatwave when your air conditioner is malfunctioning... LOL


    I personally don't think biased or unbiased makes much of a difference. I've seen just as many bad renders with unbiased renderers and good ones with biased renderers. Ultimately, it's up to the artist to tell the renderer what he/she wants. They need to understand how to coax the best results of the tools they're using.

    I'd say unbiased engines are easier to work with when you want physically correct photorealism and have access to actual scientifically obtained and calibrated parameters for your materials and lights. Then you can transfer your real-life photography know-how into 3D because it is going to be consistent with reality.
    Obviously unbiased renderers are fairly useless for those people who aren't good at real life photography (like me). I used to dabble in learning LuxRender, but it quickly began to aggravate me in the same way my real-life camera did. The only thing I really liked about it was that it was designed with HDR output in mind. I'm still trying to find any info on how to write an imager shader that would allow for rendering to a HDR format from DS... postwork-wise, tonemapping a HDR image feels more flexible to me than messing with curves'n'layers over a limited dynamic range.

    That's why I now always test with two lighting setups - one with soft diffuse light (much like photo studios) and another with harsh lights with very low UE intensity). Once you figured it out, you have a skin/materials setup that's pretty flexible regardless of lighting.

    Very low, what could it be, precisely? I usually go with 25% intensity for UE. It's probably still too high, since, as you can see, a simple change of a map changes the way UE affects scatter quite dramatically...

    Another is having both light and dark skin characters in the same space.

    This is an interesting idea, but I am afraid it won't help me because I have almost no idea how dark skin looks IRL. I live in Russia, and so if I see a dark-skinned person in the street, it's fairly rare. There are no dark-skinned people among those whom I regularly interact with. And I don't think that photos and movies are enough to get a good impression... there's a lot of editing and postproduction going on with these.

  • wowiewowie Posts: 1,987
    edited May 2013


    Very low, what could it be, precisely? I usually go with 25% intensity for UE. It's probably still too high, since, as you can see, a simple change of a map changes the way UE affects scatter quite dramatically..

    I use RGB colors 192,192,192 (to get a somewhat neutral ambient light) at 25% for the dark shot above. I didn't use any HDRI maps though.


    This is an interesting idea, but I am afraid it won't help me because I have almost no idea how dark skin looks IRL. I live in Russia, and so if I see a dark-skinned person in the street, it's fairly rare. There are no dark-skinned people among those whom I regularly interact with. And I don't think that photos and movies are enough to get a good impression... there's a lot of editing and postproduction going on with these.

    You can use other things that you are more familiar with. Candles will work since they are translucent (white for light skin, red for darker skin). Don't use the thick ones though, since you won't be able to see much scatter.

    Another tip I forgot - try rendering your materials with some color bias for UE. Adding/subtracting red and green colors (with a ratio 3:1 - 3 parts of red for 1 parts of green), while keeping the blue color intact. For example, to get a more 'cool' ambient light, I adjust my UE2 colors from 192,192,192 to 150,178,192. Light skin should be more responsive to the color changes (since they don't have as much melanin).

    Post edited by wowie on
  • Mustakettu85Mustakettu85 Posts: 2,917
    edited June 2013

    Hi everyone,

    How are you doing? Sorry I'm later than I promised... it's been a crazy couple of weeks, and pwSurface is one complicated shader =)

    Type 0 Negative, your settings are very cool! Though I felt that I could try and see if they could be made more 'predictable', so to speak, so I changed the SSS scale, turned off the ambient channel, and, on the lights, turned all the shadows on. This is why: the SSS scale of 1 makes 3Delight think the object is 10 times smaller than it is "for real", so the SSS effect becomes very heavily dependent on actual geometry; the 0.1 scale proved to be more reliable, in my experience. Then, ambient on surfaces that aren't supposed to glow in the dark can turn out to be a surprise when you decide to render a night-time scene. And as for shadows, raytraced SSS performs correctly only if all the lights cast shadows, otherwise there may be unwanted surprises, too.

    So I tried to recreate the style of your settings but with these parameters accounted for (there's quite a lot of settings that had to be changed both on the settings and on the lights, so that the looks matched the original...). Hope you like it. Check your PM for the archive with the preset files.


    What I learnt in the process is weird but interesting, and it applies to the default UberSurface and HSS as well. It's about attaching a colour map to the SSS node.

    What I'd found out before was that if you add a colour map to SSS and leave the multiplier colour at pure white (255,255,255) or a shade of grey (R=G=B), it's the same as turning SSS off or setting the colour to pure black (0,0,0). It stops working. Period.

    BUT if there is "actual colour" in the multiplier, it starts working again! The difference between the highest channel value and the lowest one should be 20 points or more for it to happen. Like (255,235,255) or (225,245,245).

    Using the correct 0.1 scale, this is enough to "tint" the SSS map without overpowering it. It opens a lot of possibilities for those shaders that only have "colour/strength" controls!


    A note on pwSurface translucence node: it is VERY different from that of omnifreaker's shaders. The "density based" mode sounds very interesting because it could potentially be useful for rendering super translucent ears without having to paint maps (and yeah, maps for ears are important for shaders that do not have "backscatter boost"). Unfortunately, it's too slow for me to properly test it right now.

    Wowie - thanks for the advice! Candles sound like an interesting idea for calibration...

    aomedit.jpg
    1236 x 800 - 124K
    Post edited by Mustakettu85 on
  • Joe CotterJoe Cotter Posts: 3,234
    edited December 1969

    ...What I learnt in the process is weird but interesting, and it applies to the default UberSurface and HSS as well. It's about attaching a colour map to the SSS node....

    This is very interesting and explains results I was getting. I had been getting the same results but hadn't fully realized what was happening, just that putting color in the color channel with the map got better results. It didn't occur to me that it was a bug where it wasn't working at all, even though it basically appeared like that. Sometimes it takes someone else explaining what you are seeing to have it actually click ;)
  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited December 1969

    Translucence?

    Render_45_3.jpg
    1920 x 1080 - 172K
  • Type 0 NegativeType 0 Negative Posts: 320
    edited December 1969

    Hi everyone,

    How are you doing? Sorry I'm later than I promised... it's been a crazy couple of weeks, and pwSurface is one complicated shader =)

    Type 0 Negative, your settings are very cool! Though I felt that I could try and see if they could be made more 'predictable', so to speak, so I changed the SSS scale, turned off the ambient channel, and, on the lights, turned all the shadows on. This is why: the SSS scale of 1 makes 3Delight think the object is 10 times smaller than it is "for real", so the SSS effect becomes very heavily dependent on actual geometry; the 0.1 scale proved to be more reliable, in my experience. Then, ambient on surfaces that aren't supposed to glow in the dark can turn out to be a surprise when you decide to render a night-time scene. And as for shadows, raytraced SSS performs correctly only if all the lights cast shadows, otherwise there may be unwanted surprises, too.

    So I tried to recreate the style of your settings but with these parameters accounted for (there's quite a lot of settings that had to be changed both on the settings and on the lights, so that the looks matched the original...). Hope you like it. Check your PM for the archive with the preset files.


    What I learnt in the process is weird but interesting, and it applies to the default UberSurface and HSS as well. It's about attaching a colour map to the SSS node.

    What I'd found out before was that if you add a colour map to SSS and leave the multiplier colour at pure white (255,255,255) or a shade of grey (R=G=B), it's the same as turning SSS off or setting the colour to pure black (0,0,0). It stops working. Period.

    BUT if there is "actual colour" in the multiplier, it starts working again! The difference between the highest channel value and the lowest one should be 20 points or more for it to happen. Like (255,235,255) or (225,245,245).

    Using the correct 0.1 scale, this is enough to "tint" the SSS map without overpowering it. It opens a lot of possibilities for those shaders that only have "colour/strength" controls!


    A note on pwSurface translucence node: it is VERY different from that of omnifreaker's shaders. The "density based" mode sounds very interesting because it could potentially be useful for rendering super translucent ears without having to paint maps (and yeah, maps for ears are important for shaders that do not have "backscatter boost"). Unfortunately, it's too slow for me to properly test it right now.

    Wowie - thanks for the advice! Candles sound like an interesting idea for calibration...

    I simply can’t thank you enough for this. Your work looks great and I can’t wait to try it out. Feel free to do what you want with this, you may share it on sharecg if you wish. I’ll take mine down soon. I hope to get some free time for DAZ Studio later this week.

    pwSurface density based translucence is really neat, I tried it out a bit but it is slow, so slow i gave up on it but got some interesting looking things to happen, kind of weird on skin though, more like it would be good for glass-like things perhaps. Could possibly be used on teeth and eyeballs, that would not hurt the render time so bad and might be usefull. Hope to have some time for that idea soon.

  • wowiewowie Posts: 1,987
    edited December 1969


    What I'd found out before was that if you add a colour map to SSS and leave the multiplier colour at pure white (255,255,255) or a shade of grey (R=G=B), it's the same as turning SSS off or setting the colour to pure black (0,0,0). It stops working. Period.

    BUT if there is "actual colour" in the multiplier, it starts working again! The difference between the highest channel value and the lowest one should be 20 points or more for it to happen. Like (255,235,255) or (225,245,245).

    Using the correct 0.1 scale, this is enough to "tint" the SSS map without overpowering it. It opens a lot of possibilities for those shaders that only have "colour/strength" controls!



    This is very interesting and explains results I was getting. I had been getting the same results but hadn't fully realized what was happening, just that putting color in the color channel with the map got better results. It didn't occur to me that it was a bug where it wasn't working at all, even though it basically appeared like that. Sometimes it takes someone else explaining what you are seeing to have it actually click ;)

    I believe the upper limit is anything above 250,250,250. Values below that works fine.

  • Mustakettu85Mustakettu85 Posts: 2,917
    edited June 2013

    Translucence?

    Hi there! =) How did you get the light to shine through that belly?

    wowie said:

    I believe the upper limit is anything above 250,250,250. Values below that works fine.

    I didn't test with HSS, but both the default US and pwSurface2 consistently refuse to activate SSS when R=G=B, even if we're way below 250.


    pwSurface density based translucence is really neat, I tried it out a bit but it is slow, so slow i gave up on it but got some interesting looking things to happen, kind of weird on skin though, more like it would be good for glass-like things perhaps. Could possibly be used on teeth and eyeballs, that would not hurt the render time so bad and might be usefull. Hope to have some time for that idea soon.

    Teeth sounds interesting, if they're modeled with enough detail it might be very impressive, they should be more translucent around the sharper edges then. IDL should probably be enabled on the inner mouth, too, for there to be more light to pass through them.

    Post edited by Mustakettu85 on
  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited December 1969

    Translucence?

    Hi there! =) How did you get the light to shine through that belly?

    Hi Muskatt, thanks for the word on my DeviantArt account. How have you been doing ? That's been a long time since we had a discussion in a thread : ). The answer to your question is Point Light inside the body of course. That's one of the interesting effect of SSS

    I should give advices here but there are so many I don't know where to begin

    I don't know either if I should post examples with shaders I wrote to illustrate my point

    However here are a few things :
    - render with point cloud script if you can. That speeds up things
    - people tend to confound SSS and softness due to specular with low glossiness. Put a back light to see the real effect or minimize specular effect
    - begin with the basics and render without maps. Skin is very complex
    - I don't think US, US2 or PW Surface are physically plausible shaders so don't bother too much about energy conservation or any other law.


    Here are some simple renders with red/pink SSS and the use of double specular to simulate coating over the SSS layer. Choosing a good color for the first specular layer can give interesting results. Only UE + US + a distant light with raytraced shadows

    Render_20_13.jpg
    557 x 564 - 131K
    Render_19_13.jpg
    607 x 571 - 109K
    Render_18_14.jpg
    609 x 617 - 110K
    Render_17_17.jpg
    610 x 605 - 95K
  • Mustakettu85Mustakettu85 Posts: 2,917
    edited December 1969

    The bunnies look superb, Takeo. Where are those old test meshes (the bunny, the dragon, the teapot) available from now?

    I don't know either if I should post examples with shaders I wrote to illustrate my point

    I'd say you should... if anything, it may inspire others to study the possibilities of shader builder...


    - begin with the basics and render without maps. Skin is very complex

    For me, actually adding maps is the most complicated thing! I still plan to get that procedural shaderMixer material done... anything to avoid maps, LOL


    - I don't think US, US2 or PW Surface are physically plausible shaders so don't bother too much about energy conservation or any other law.

    It's still a good way to avoid "overexposed" surfaces, in the first approximation - particularly when using the diffuse map in the diffuse channel. I don't do that anymore with US2 (I stick desaturated maps in the diffuse), and that latest discovery that the default US can also take the colour map in the SSS colour multiplier means that I should experiment with the same technique for it... that way sliders could be rather arbitrarily turned up to 11, LOL

  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited June 2013

    I'm doing test with Stanford Scan available at http://graphics.stanford.edu/data/3Dscanrep/ . These are .ply high res mesh and you can use Meshlab (free) to convert to obj.

    You can also download obj. files for DS here http://graphics.cs.williams.edu/data/meshes.xml

    I also play with Sibenik Cathedral to test some lights I made. There are very good models there

    I played a lot without texture to understand how US works because there is one major problem with texture is that it is already the resulting color of light scattering through skin which is 3 layer, each with different properties. So that won't be as easy. To solve this problem somebody from Pixar had the idea to make a big table to determine which dmpf and scattering color should be used depending on the diffuse texture color. That's a bit more complicated as just specifying one scatter color for the overall skin and wanting to get something realistic

    Plus I've seen a study where they mesured the diffuse reflectance on a face and the results showed that the reflectance was hugely varying over just the face and was not as predictable as the simple dipole SSS approximation from Jensen's.

    That is why I say that one should know how the shader works and fake things. That is how it goes. We can only approximate reality

    The most accurate way would be to use Luxrender or some other unbiased engine (eventually Octane but I'm reserved as to it's accuracy because I think they cheat to get results faster)
    Luxrender uses accurate bssrdf and spectral light to do calculations with monte carlo algorithm and russian roulette if I'm not mistaken. That's what you don't have with our current 3delight rendering. But even though an accurate way of calculating is yet to be implemented and I didn't see any. Just doing volume rendering with same properties on the whole volume and using a diffuse map is still not accurate. That is why you still have to cheat or you implement a table in which you lookup for each color of the diffuse map. And I'm not even talking about making an algorithm which detects if it is skin or not and have different parameters for make up or anything that is not skin on the diffuse map. Doing something really accurate would ask lots of calculations and I don't think people would want to wait to get it. So cheating is the best thing to do.

    As for procedural skin shader I'm not sure it is a good idea because it will be a bit monotonous. Using a map is still the best way to break it. But getting a good "cheating" calculation is something not easy.

    Now for physically plausible shading. Pixar implemented it in it's last release of Renderman (that I'd wish to have)

    Have a look at the videos here http://www.youtube.com/watch?v=b8LFBX4x4qE and http://www.youtube.com/watch?v=Z8R0v0HO9Io

    It will explain also a lot why we often see crappy renders
    We don't have that in 3delight and I don't know how much time they will need to get to that or if it will ever be implemented. I began to test RSL 2.0 programming because I'd like to get faster renders on transmapped hairs but 3delight has still many thing to implement to get there.

    About the different shaders you have in DS It may be interresting to test every shaders with the same scene and light and see how they each react to eventually see the differences. I only have US so I can't test US2 or PWSurface. US2 should be better than US

    Finally as for inpiring picture I don't know. I made so many test renders (just looked at one folder : 1183 files 789 Mo) Do you like rabbits ? (PS there is a point light behind the green rabbit that is why it glows)

    Render_9_25.jpg
    1920 x 1080 - 1M
    Post edited by Takeo.Kensei on
  • Joe CotterJoe Cotter Posts: 3,234
    edited June 2013

    Thank you very much for this post, it has a lot of good information.

    As for the cheating, I tried to explain this before to people who were trying to force some real world calculations on to shading and they just weren't getting it. I of course did not have the level of information you provided here... I was coming from a background of programming and art, where cheating is the word of the day sometimes, and that alone wasn't convincing ;)

    The basic concept of putting in some test data and seeing if one gets consistent results still applies. If we don't... we either need to update our methods (in this case, the underlying render engine or at least shader code) or conversely accept that the process cannot be automated with consistent results. The last stage is of course deciding if the results we do get are within acceptable parameters or, if they can work to cut the amount of effort required to hand adjust the process to achieve the desired results.

    Post edited by Joe Cotter on
  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited December 1969

    No problem Gedd. I know that people complain about DS documentations but frankly most part of the knowledge can only come from scientific papers
    I'm sure almost nobody would read what I've read so far to understand all that (lots of math and physics)


    There is also the fact that Pixar's renderman (and thus every renderman compliant renderer) is not designed for realism first but mainly to be easily customisable, controlable, scalable and versatile and that you normaly need an army of RSL programmers to get what you want (or should I say the cheating look that you want). As is, rendering in DS is limited because there are barely few shaders. As for Omnifreaker's shader, I think he made a good Ubershader that is able to approximate most materials but you may need specialised shaders if you want something better. For example you can't use it for diamond shading with chromatic abberration and if I want that I'll have to implement it myself knowing the fact that I need to do calculations for each wavelenght with a spectral light that we don't have. So faking or not faking?

  • Joe CotterJoe Cotter Posts: 3,234
    edited December 1969

    Well the other thing is that generic shaders have both limitations and a lot of settings that are unnecessary for many uses one might put them to. There is potentially a vast market for shaders in DS/3DLight that would both specialize and simplify defining surface properties.

  • wowiewowie Posts: 1,987
    edited June 2013

    Thanks for those. I think they are very helpful.

    Edit: I watched both videos and I think although we won't get physically accurate materials without those shaders, we can have a reasonable approximation. It won't be as flexible or as foolproof.

    Below is a render of my new settings, mostly based on getting both SSS and specular as close as possible with Elite Lana. I saved the materials and applied that to M4's with the default texture (only diffuse and bumps were used, in their respective slots). I did have to make SSS strength maps for all the skin surfaces. The diffuse values in the color sliders most probably are well off (since if you want proper SSS you need to desaturate the texture, either in an image editor or by cancelling out the color).

    Looks rather nice (at least I think so).

    M4.jpg
    800 x 1040 - 213K
    Post edited by wowie on
  • j cadej cade Posts: 2,256
    edited December 1969

    I'm doing some thread necromancy I suppose (Hey its better than thread necrophilia!) But I gave been messing with shader mixer and have made something rather cool. The relevant surfaces were made using 4 images: a bump map a spec map and two black and white control textures. specular/reflections and opacity are fully controllable for each colour and the displacement is sort of based on the specular.

    I do have a question for anyone who can help. Is there any way to control the order things show up in the shader panel? Also some categories refuse to be hidden, they just go grey, is there any way to fix this?

    demo5b_copy.jpg
    800 x 1200 - 370K
    demo1.png
    500 x 500 - 451K
    demo2.png
    500 x 500 - 530K
    demo3.png
    500 x 500 - 543K
    demo4.png
    500 x 500 - 481K
  • Joe CotterJoe Cotter Posts: 3,234
    edited December 1969

    Don't have answers for your question atm, but the results look nice :)

  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited December 1969

    Kamion99 said:
    I'm doing some thread necromancy I suppose (Hey its better than thread necrophilia!) But I gave been messing with shader mixer and have made something rather cool. The relevant surfaces were made using 4 images: a bump map a spec map and two black and white control textures. specular/reflections and opacity are fully controllable for each colour and the displacement is sort of based on the specular.

    I do have a question for anyone who can help. Is there any way to control the order things show up in the shader panel? Also some categories refuse to be hidden, they just go grey, is there any way to fix this?

    Very nice.

    Text editing is my favorite weapon for what you ask. Save the shader as dsa preset then edit it.
    There are however some parameters that should stay untouched depending on your network. Can't say much more without a network in front of the eyes

  • wowiewowie Posts: 1,987
    edited December 1969

    Reworked, adding some SSS strengh maps and a poor attempt at hemoglobin map (just to add a bit of redness here and there).
    Also tweaked the gamma and gain in the render options. So, please if they looked washed out or not.

    Thanks.

    Render_14.jpg
    800 x 1040 - 375K
Sign In or Register to comment.