3Delight Laboratory Thread: tips, questions, experiments

13334363839100

Comments

  • Thank you Zarcon. I thankfully don't drive, we have good public transportation system in this city =)

    The thing with the vases is that actually the Renderman standard wasn't ever specifically optimised to deal with polygonal meshes - and I don't think the 3DL guys ever bothered ("pure" meshes are not that efficient). Going extra-hi-poly isn't the perfect solution for this reason; a subdivision surface will always work better, and SubD's are among the things the renderer works with best. "Smoothing" is not the same as using a SubD - as you can see it will fail, especially when normals are disturbed.

     

     

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited November 2015

    yea, I kind of live out in the wilderness. Public transportation consists of a pair of snow shoes, lol.

    I tried sub-D on that vase, I cranked it out to three (the max), and it did nothing to those bands. I discover some obscure switch in the Omni shader that mended it some, "Trace Displacements" on (I'm Still not sure exactly why the switch is there, only that it helps the smoothing out some). I'm still trying to figure out what a few settings are supposed to do, as I slowly work my threw a thread.

    http://www.daz3d.com/forums/discussion/30047/ubersurface-tutorial

    Ah, never mind. I can't tell if that is for US2 or omUberSurface(included in studio). I'm taking five from this, it's already been a very long day. I need to draw up a list, as I'm just not seeing the answers for a few things.

    Post edited by ZarconDeeGrissom on
  • I justified the purchase of a 980 Ti in part with Iray, but after getting the card I've only really been interested in 3DL. Funny how that happens sometimes. Pretty sad to see that DAZ doesn't seem to care about 3DL much anymore.

    Anyway, I've been skimming this thread again (the technical stuff is mostly beyond me) and see you often mention Kettu's lights and shaders. What are they, and can I have them? blush Are you talking about the ones linked in Kettu's signature?

    Also if there is something I can do to help, like pester DAZ tech support with 3DL bug reports and such, please let me know.

     

  • mjc1016mjc1016 Posts: 15,001

    The lights and shaders some are in her sig, the others are still in testing and you'll need to drop Kettu a PM/email...

  • mjc1016mjc1016 Posts: 15,001
    edited November 2015

    Another set of raycache on/off renders... the 'off' comes in at 45 mins, the on at 4.5 mins.

    Off:

    (the sleeves are my fault...)

    On:

    lights_1.jpg
    640 x 640 - 106K
    lights_2.jpg
    640 x 640 - 102K
    Post edited by mjc1016 on
  • Mustakettu85Mustakettu85 Posts: 2,933
    edited November 2015

    Zarcon, all those extra sub-d levels are meaningless in 3Delight - it always renders the 'limit surface' (maximum quality - as long as it's a subdivision surface and not a 'base mesh'). But if that vase is using displacement, then of course it should be included in raytracing, otherwise the result will be wrong (in various degrees, depending on how much raytracing is involved - if you use the raytrace hider, like in 'progressive rendering', it's mandatory if you want to see your mesh actually change shape).

    Turning raytracing the displacement off used to save resources years ago. Not so much these days. The DS default shader always traces it.

    The question is, though, WHY displacement for this fine surface relief?? It's an 'oldschool' shader plus I couldn't see the displacement changing the shape of the vase. Bump mapping should be enough in this case.

     

    Post edited by Mustakettu85 on
  • Araneldon,

    Yup, as Mjc says, to get the most interesting stuff you need to send me a PM because the kit is still in the 'alpha' stage - which means documentation is unfinished and there are not many handy/user-friendly presets. The actual shaders and render scripts are fairly robust, though.

    As for my freebies, what I have on ShareCG is 'oldschool' but still useful in part when coupled with the alpha kit: the 'Fantasy Lights' come with 'non-realistic' HDR maps of alien/fairytale skies, and the subsurface scattering tutorial explains the basics of how SSS is dealt with in 3Delight, which don't change much with the introduction of new features that I use in my shader. 

  • wowiewowie Posts: 2,029
    edited November 2015

    Adding to Mustakettu's input:

    You only need displacement (with raytracing) if you want displacement details to be seen in the shadows and/or silhoutte. Well, at least in (non realtime) CPU renderers. There are hacks such as relief mapping and parallax occlusion, which is typically used in real time game engines.

    In layman's terms it's the difference between using a wallpaper with images of bricks and using actual bricks on the wall.

    figure4-25-right.jpg
    260 x 288 - 83K
    figure4-25-left.jpg
    260 x 288 - 82K
    normalvsdisplce_displaced2.jpg
    624 x 310 - 49K
    Post edited by wowie on
  • Thank you wowie and kettu. I had just tossed it there as a random choice for the ones I was not using the 'Normal' maps on (normal on the left vase).

    My difficulty was with the Velvet and Gloss settings. I will look at it again after I head to the Post office to send out the Disability reevaluation question thing (BRB). It's been a stressful two weeks trying to remember stuff from years ago. I'm sure my mind was not in the best state to comprehend allot of new stuff.

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited November 2015

    OK, no offense to Scott Livingston. I needed a tad more then some of the "Definitions" of some of the dials, and equally as important was the 'Direction' of each dial.

    I'm still not sure it will ever completely soak in to my brain, tho at least I now have some visual examples to compare so I can see what direction to move what dial to get more or less of something. And possibly it may be useful to others that are more of a 'Visually' learner rather then auditory in a way. I know there is allot of stuff I didn't go over (and don't know), as it is more of a fundamental Basics of a sort.

    http://www.daz3d.com/forums/discussion/comment/947916/#Comment_947916

    Post edited by ZarconDeeGrissom on
  • wowiewowie Posts: 2,029
    edited November 2015

    Hi Zarcon,

    UberSurface's Specular Glossiness and Roughness are the same thing. Technically, Roughness is the correct term nowadays. Glossiness is the older term for it.

    These are tied to diffuse - Velvet, Translucency and Subsurface Scattering. Diffuse roughness affects them too.

    As for specular color - generally only metal have 'colored' specular and reflection. Specular and reflection for non metals (dielectriic) are always white (or very close to it). And as a side note, always put specular maps in the strength slot. Using specular maps in the color slot and changing specular color are just really bad habits.

    One of the most important feature of UberSurface/UberSurface2 that is generally overlooked is the Occlusion Shading Rate Mode override. Make sure to set it to Override and change Shading Rate to 32 or above for hair or any other surfaces that makes heavy use of opacity maps.

    Fantom set to On makes the surface/object invisible to the camera, but may still cast shadows and reflections.

    Raytrace set to Off makes the surface/object invisible to reflections and refractions. Assuming you're doing raytraced reflections/refractions on other objects/surfaces.

    Accept Shadows and Occlusion should be pretty explanatory. Set to On if you want the object/surface to receive shadows and Occlusion.

    Fresnel only affects reflection, not specular. The higher the strength, the less reflection you'll see when viewing the surface head on. When viewed at glancing/grazing angles, reflection is stronger. Falloff and Sharpness controls the rate from the edges to the center.

    Post edited by wowie on
  • I have seen "glossiness" used lately - some Iray manual for some other software, I think - but technically it's the opposite to "roughness" (although Omnifreaker used these as synonyms). A glossiness of 1 should be the same as a roughness of 0 (both physically implausible values for a contemporary shader) - a perfectly polished material.

     

  • wowiewowie Posts: 2,029
    edited November 2015

    And at 1 (total glossiness or zero roughness), you typically don't have any specular, since a perfectly polished material is basically a perfect mirror.

    Myself, I prefer using roughness, since you can also use it for diffuse and reflection/refraction (instead of 'Blur').

    So, I've finally got a really nice script based of Casual's mcjspreadtex script (which I asked him to write). In addition to copying textures (diffuse, bump, displacement) from the 1st channel to the 2nd, I've modified the script so it also copy colors from the 1st channel diffuse to the translucency and 2nd diffuse, normalized to a HSV value of 0.6275 (or a max value of 160 in RGB scale). It also clamps low values to 32,32,32 since there is no pure black in real life.

    http://www.daz3d.com/forums/discussion/65526/copying-modifying-surface-color-values#latest

    Here's a shot of the skin MAT preset. I've tinted the diffuse a bit on both. Generally, both are using the same preset (specular, fresnel, SSS) with slightly modified diffuse values and different specular strength since diffuse textures and specular maps from vendors vary wildly from one set to the other.

    With a more ambient lighting setup. Same presets.

    Still need to make thumbnail icons for some stuff and update my universal MAT/shader presets, but I (finally) think it's finished. I'll be making a manual and a no-bullshit guide to go along with it.

    1.jpg
    823 x 1070 - 378K
    2.jpg
    823 x 1070 - 362K
    3.jpg
    823 x 1070 - 380K
    4.jpg
    823 x 1070 - 358K
    Post edited by wowie on
  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited November 2015

    mjc1016, that is a cute one, I like the ears. As for the Ray caching, I'm not seeing much difference at all (that I can tell with just my eyes).

    wowie, the work on those skin mats is really good looking. Outstanding compared to my pathetic AltShader attempts, lol. I know, completely different purpose, methods, shaders, and goals between them.

    Roughness, I figured it should somehow be adding noise or irregularity to the effect some how, lol.  Roughness; having a coarse or uneven surface, as from projections, irregularities, or breaks; not smooth. I guess you can understand my confusion with that, lol.

    As for the rest of the Uber Shader stuff, I'm still fumbling threw it, lol.

    Mirrors vs Glossienss, and a good Q. What is the Roughness or Glossienss of a Telascope mirror, like the ones at VLT? It must be far beter polished then the one in my bathroom, lol.

    http://www.eso.org/public/usa/teles-instr/vlt/

    Post edited by ZarconDeeGrissom on
  • mjc1016mjc1016 Posts: 15,001
     

    Roughness, I figured it should somehow be adding noise or irregularity to the effect some how, lol.  Roughness; having a coarse or uneven surface, as from projections, irregularities, or breaks; not smooth. I guess you can understand my confusion with that, lol.

    As for the rest of the Uber Shader stuff, I'm still fumbling threw it, lol.

    Mirrors vs Glossienss, and a good Q. What is the Roughness or Glossienss of a Telascope mirror, like the ones at VLT? It must be far beter polished then the one in my bathroom, lol.

    http://www.eso.org/public/usa/teles-instr/vlt/

    It is but not at a 'macro' level...it's using math to say that the surface is not perfectly smooth, as opposed to a noise map or something you can actually see.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited November 2015

    Great work, Wowie! Bring on the no-bullshit stuff =)

    It would be awesome if DS image editor could be easily programmed to adjust texture levels so as to "unblack" textures automagically, not just multiplier colours. It can be done in the shader, but it's not exactly a "free" operation computation-wise.

    -----

    Zarcon, here are two pages that might help you visualise how math is used to account for roughness:

    http://simonstechblog.blogspot.ru/2011/12/microfacet-brdf.html - if you scroll through the densest math forest, there will be an applet to play with!

    http://www.codinglabs.net/article_physically_based_rendering_cook_torrance.aspx

    Post edited by Mustakettu85 on
  • wowiewowie Posts: 2,029
    edited November 2015

    Great work, Wowie! Bring on the no-bullshit stuff =)

    It would be awesome if DS image editor could be easily programmed to adjust texture levels so as to "unblack" textures automagically, not just multiplier colours. It can be done in the shader, but it's not exactly a "free" operation computation-wise.

    I've been doing some thinking about that. I'd say the biggest problem with DS are textures - be it color/diffuse or control maps. For control maps, be it opacity and specular, we should probably disregard any value other than 0 or 1. Effectively making them masks. Would easily solve all problems with specular strength.

    As for the diffuse, implementing clamping via the shader much like the script I've modified would be a good start. But as you said, it doesn't come freely. The highest value should be set by the specular (ie diffuse = 1 - specular & reflection strength). With 1 being total reflectivity. That should make it plausible and there's no need to do fancy IOR stuff like that Arnold shader.

    Zarcon, here are two pages that might help you visualise how math is used to account for roughness:

    http://simonstechblog.blogspot.ru/2011/12/microfacet-brdf.html - if you scroll through the densest math forest, there will be an applet to play with!

    http://www.codinglabs.net/article_physically_based_rendering_cook_torrance.aspx


    Hell, he should read this in that second website you've linked.

    http://www.codinglabs.net/article_gamma_vs_linear.aspx

    Post edited by wowie on
  • Slowly working my way threw a few cans of worms here, lol. Are they working on that Raycaching, or did they make a mistake.

    P.S.

    Squashing or Expanding blacks. An odd counter argument (tho just hypothetical in origin, don't shoot me, lol), regarding decreasing the depth of black in an image (not gama, just raising the minimum black value). CRTs can do black, real deep black, so I would agree with you for my G220fb CRTs. Flat panel monitors that most can afford can not do a deep black, thus for some it may make things look worse on there screens to raising the minimum black value. Just something to consider.

    20151118_Odd_Stuff_3DL_002.png
    1536 x 738 - 188K
  • wowie said:

    For control maps, be it opacity and specular, we should probably disregard any value other than 0 or 1. Effectively making them masks. Would easily solve all problems with specular strength.

    I'm not sure that two-bit masks would be always fine. Think of a feather - you'd want gradients of opacity to make a realistic one. Same with hair, actually. I remember having to deal with two-bit opacity masks on some hair in Oblivion - that looked noticeably worse than 8-bit gradients.

    Or if you have metallic decorations painted on a fabric texture, and if the fabric is also a fancy one - then there should be grades of specular, too.

    wowie said:
    Hell, he should read this in that second website you've linked.

     

    http://www.codinglabs.net/article_gamma_vs_linear.aspx

    Everyone who's still not getting linear workflow should read it. And all the other articles on the subject, too. Until it sinks in.

  • CRTs can do black, real deep black

    Yes, but the point is - in your render, blacks should come from absense of light. Not unphysical spots on a texture that just won't reflect light because someone painted them RGB (0,0,0). The colours are multipliers to the incoming light. No colour = no light.

    DS linearises colours with GC on, so Wowie's (32,32,32) bottom limit is just 1% of total incoming light. That's quite black, but it's way more real-world black than (0,0,0). The latter is a veritable black hole.

    Dark wet soil reflects 5%: http://www.eoearth.org/view/article/149954/

  • mjc1016mjc1016 Posts: 15,001

    Vantablack...a carbon nanotube material absorbs 99.86% of the light hitting it...it's so black it can't be 'seen'.

    The next step up, carbon black is in that 1-2% range.  There are very few things that one would use carbon black as a color for.

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited November 2015
    wowie said:

    For control maps, be it opacity and specular, we should probably disregard any value other than 0 or 1. Effectively making them masks. Would easily solve all problems with specular strength.

    I'm not sure that two-bit masks would be always fine. Think of a feather - you'd want gradients of opacity to make a realistic one. Same with hair, actually. I remember having to deal with two-bit opacity masks on some hair in Oblivion - that looked noticeably worse than 8-bit gradients.

    Or if you have metallic decorations painted on a fabric texture, and if the fabric is also a fancy one - then there should be grades of specular, too.

    wowie said:
    Hell, he should read this in that second website you've linked.

     

    http://www.codinglabs.net/article_gamma_vs_linear.aspx

    Everyone who's still not getting linear workflow should read it. And all the other articles on the subject, too. Until it sinks in.

    And just how much, "sinking" (lol), is it going to take,

    befor I can see what I'm doing with GC on?

    And When am I going to get a GC on,

    that dose not wash out all the colors?

    Been there, tried that GC-on junk. It dose not work, in daz studio with 3DL. As far as I'm concerned, the discussion is over. If the drive for the Laws of Gama Correction must be on is like the rules of physics that are supposed to apply everywhere, well you can't have two separate everywheres. lol.

    And for those that don't know, this is how that render should have looked.

    zdg_Graviton sVsSpacetime.jpg
    500 x 374 - 41K
    Post edited by ZarconDeeGrissom on
  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited November 2015
    mjc1016 said:

    Vantablack...a carbon nanotube material absorbs 99.86% of the light hitting it...it's so black it can't be 'seen'.

    The next step up, carbon black is in that 1-2% range.  There are very few things that one would use carbon black as a color for.

    Yea, and that is part of why I have doubts about limiting the color depth of a map, especially if the map was from a Photo reference with a real world black limit already in it.

    Tho good point Kettu regarding the lights in a render casting the 'Black'. Most things I use between 11 and 32 (24bit RGB) for a black, except that color scale cube (black on that is 0r 0g 0b). Just consider the black on the color scale cube a black hole or carbon nanotubes, lol.

    Post edited by ZarconDeeGrissom on
  • Zarcon, dear, if only you would use different phrasing, what you say would be way less misleading to lurkers.

    "It dose not work, in daz studio with 3DL" - this is plain not true, as quite a few of those who use 3Delight in DS for real will testify. It works, and beautifully.

    You dislike the way stock materials look with GC on? That's right; materials and lighting need to be done differently for linear workflow. And the other way around.

    As for your viewport, I don't really get what's wrong with it?

    Of course, if you're so inclined, it's your choice to reject whatever you want... but it won't change the facts in general.

  • wowiewowie Posts: 2,029
    edited November 2015

    I'm not sure that two-bit masks would be always fine. Think of a feather - you'd want gradients of opacity to make a realistic one. Same with hair, actually. I remember having to deal with two-bit opacity masks on some hair in Oblivion - that looked noticeably worse than 8-bit gradients.

    Or if you have metallic decorations painted on a fabric texture, and if the fabric is also a fancy one - then there should be grades of specular, too.

    Most of the problems with opacity masks are due to improper alpha blending. Funny you should mention a game like Oblivion. With realtime graphics, the only way to anti alias opacity textures is via super sampling. That means higher resolutions maps (more points) rather than gradients.

    If you still want a slightly transparent look to it, just use a non mappable opacity strength. The only drawback I see with such an approach is mixed materials - which I think is a bad idea anyway.

    Zarcon, your light setup is strange.

    50% strength with a HDRI? I generally see intensity needs to be twice if you use a HDRI. So that makes your ambient light around 25% (with pure white). I ended up with this when I render with those settings.

    The plane and sphere is set to pure white. Of course, this is with gamma correction enabled. If you add any light (spot, distant or point), assuming you're using the same pure white color, the maximum intensity of any other light in the scene needs to be very low (somewhere around 70-75%, distant light or no falloffs).

    Converting the values to my setup, I ended up with UE2 at 130.0% with 192,192,192 and a max intensity value (for any other light) of 500.0% with 192,192,192. This is with a material color of 160,160,160. The ambient light intensity is generally about 1:2.5 what I normally use.

    5e7a3d787b9b9f34a32784157740d4.png
    356 x 594 - 22K
    Z1.jpg
    1070 x 1070 - 258K
    Post edited by wowie on
  • wowiewowie Posts: 2,029
    edited November 2015

    This is how my viewport looks like with your light setup, converted into mine. Everything is generally quite visible.

    Viewport.jpg
    1920 x 1160 - 395K
    Post edited by wowie on
  • wowie said:

    Most of the problems with opacity masks are due to improper alpha blending. Funny you should mention a game like Oblivion. With realtime graphics, the only way to anti alias opacity textures is via super sampling. That means higher resolutions maps (more points) rather than gradients.

    I'm a bit rusty on the details, as I haven't done any major Oblivion modding in years - but Gamebryo (Oblivion engine) did support gradient alphas; I'm sure it's commonplace these days. And you could control the falloff curve from 0 to 1: http://niftools.sourceforge.net/forum/viewtopic.php?f=16&t=2828

    I think I had some DDS tool that only saved out the on-off alpha mask, not a gradient (DXT3 compression, not DXT5). And that annoyed me to no end because I couldn't get the more realistic look to my hair transparencies.

    So, have you been able to get nice-looking hair with a strict b&w opacity in 3DL?

  • wowiewowie Posts: 2,029
    edited November 2015
    So, have you been able to get nice-looking hair with a strict b&w opacity in 3DL?

    I ran the Elite hair opacity maps through GIMP's threshold, with a value of 1 and 254. You can see the difference between the original texture (left) and the resulting one (right).

    Here's a render of the hair with both.

    Unedited

    Edited

    One note though, doing it this way means a lot of opacity maps needs to be re-edited (anti aliased, leftover brush strokes removed, levels normalized/equalized).

    Test.jpg
    1375 x 694 - 510K
    2.jpg
    1070 x 1070 - 624K
    1.jpg
    1070 x 1070 - 614K
    Post edited by Chohole on
  • Thanks a lot for the example! I think the unedited version looks more "natural". The edited version gives the strands a stiff, "gelled and ironed" look. Of course, it could be exactly what the artist wants =)

    Or maybe there's only one person in a thousand who will notice the difference. I do, but?..

    What makes me sad is that there are still diffuse maps being released for new hair models that have highlights burnt in... like the new hairstyle that comes with G3M. There must be something wrong with my install, though - the 3DL mats that load for it don't make sense at all, it's like they were Poser mats files (DS default shader, only loading maps and no parameter values).

     

  • mjc1016mjc1016 Posts: 15,001

    What I think looks best is a compromise map...throw a little Gaussian blur (just two or three pixels worth).  It softens the harsh edges without being too 'muddy'.  One of the problems with the edited map is you end up with a lot of not very smooth edges stemming from jpg artifacts...it would probably work a lot better on an uncompressed lossless format for the image file.

    And that is one of the things that really bugs me...in this day and age using jpgs for the image files makes no sense.  There are just too many quality losses with them...and the renderers are so much better, the mesh resolution is better, but even at high resolution jpgs still suck.

Sign In or Register to comment.