3Delight Laboratory Thread: tips, questions, experiments

16667697172100

Comments

  • wowiewowie Posts: 2,029
    edited December 2016
    dracorn said:
     

    Thanks wowie and ZarconDeeGrissom.

    I am using Daz Default Shader, changed the background to blue sky (there was only an environment backdrop before), and lowered the reflection to 25%.  Both the specular and relfection are neutral gray and white, respectively.  I am still getting the blue line and no reflection on the eye, and frankly, I'm stumped.  Here are my settings:

     

    From the looks of it, it's probably the reflection. Try turning that off (the refraction too, just in case) and stick to just specular for the highlights. If you really want to have reflection, use UberSurface.

    By the way, have you checked if 'Raytrace' is enabled in the environment ball surface? If it's disabled, it won't be visible to reflection, but still show up in the viewport. By default, it is disabled.

    And of course, the ball should fully envelope your scene and be the parameters should be set to 'Visible in Render'. That's also disabled by default.

    Environment Ball.jpg
    337 x 498 - 45K
    Environment Ball2.jpg
    535 x 236 - 34K
    Post edited by wowie on
  • *Whistles a tune while idly contemplating new render engine plugin ideas*

    Basic 3Delight or Renderman 21... Decisions, decisions...

     

  • LotharenLotharen Posts: 282

    I have a question. How do you light a large scene? While I have been able to light a single model I'm lost at how to accuratly light a large scene. Specifically an indoor scene (would love to know for an outside scene too).

    I've been using iray to much and am honestly lost when it comes to 3Dlight. Any tips/information would be greatly appreciated, and thanks in advance!

  • Oso3DOso3D Posts: 15,045

    So I bought Wowie's Lumina thing ( http://www.daz3d.com/lumina-materials-library ) and am trying to make sense of it.

    It has 'bounce lights,' but as near as I can make out, all they are are ambient light sources.

    Am I missing something, or is this some broader, looser definition of bounce light?

     

  • Oso3DOso3D Posts: 15,045

    ... Did something change with Atmospheric Cameras even more?

    I could swear there was a 'only apply lights flagged with foglight category' but... either I'm using the wrong cameras or the function was just completely removed.

     

  • wowie said:

    I stopped doing work on it a couple of months ago, after that Houdini sneak peak. FEM muscles laughsmiley Oh, that's just so nice.

    I applaud you for trying. I see those muscle/skin rigging systems, and I just can't tell myself to go and rig something weight-mapped. Hell, I can't even convince myself I may need it for gamedev one day because I sure as hell am not going 3D before Jeff Vogel of Spiderweb does.

  • ... Did something change with Atmospheric Cameras even more?

    I could swear there was a 'only apply lights flagged with foglight category' but... either I'm using the wrong cameras or the function was just completely removed.

    Only the EasyVolume cam has it. Others are shader mixer and so cannot use categories.

  • So I bought Wowie's Lumina thing ( http://www.daz3d.com/lumina-materials-library ) and am trying to make sense of it.

    It has 'bounce lights,' but as near as I can make out, all they are are ambient light sources.

    Am I missing something, or is this some broader, looser definition of bounce light?

    The "oldschool", hand-placed definition. Do you have a library nearby that has Birn's "Advanced lighting and rendering"? It's expensive to buy, even on Kindle, but it's worth reading at least once.

  • Lotharen said:

    I have a question. How do you light a large scene? While I have been able to light a single model I'm lost at how to accuratly light a large scene. Specifically an indoor scene (would love to know for an outside scene too).

    Indoors are somewhat trickier. Basically with what DS currently has, your best bet would be to use an "oldschool" approach, with manually placed lights (points or spots) wherever you need an accent like colour bleeding / bounce. You can get Wowie's kits here in the store and take a look.

    For outside scenes, you could check out my free "fantasy lights" set (see the link in my signature for the freebie thread). It comes with some sort of explanation in PDF format. But generally, they are easier to work with.

  • Oso3DOso3D Posts: 15,045

    Only the EasyVolume cam has it. Others are shader mixer and so cannot use categories.

    Ah.

    Yeah, all the volume controls vanished. Then I got suspicious, reinstalled... and they're back. Whew.

     

     

  • Oso3DOso3D Posts: 15,045

    And now 'cast shadows' US2 flag isn't working with it, all of a sudden.

     

    (facedesks a few times)

    You know, 90% of why I moved to Iray was this constant being jerked around by 3DL's unpredictable behavior.

     

  • Oso3DOso3D Posts: 15,045

    Going to make 3dl work for me if it kills me...

     

    So, again, occlusion and hair... what should I do? I have it set to US2 and Occlusion: off, but it's still taking a hideous time to render.

     

  • And now 'cast shadows' US2 flag isn't working with it, all of a sudden.

    You know, 90% of why I moved to Iray was this constant being jerked around by 3DL's unpredictable behavior.

    Will, this is pure DS being its unpredictable self, not 3Delight. Those flags are set by DS rendertime scripts, not shaders even.

    There's a number of those scripts working behind the scenes. They are more reliable than shader mixer autogenerated ones, but still - reinstalling should help.

     

    So, again, occlusion and hair... what should I do? I have it set to US2 and Occlusion: off, but it's still taking a hideous time to render.

    If you are using soft raytraced shadows on your other lights (of course you are =)), make sure you're using "progressive mode" to get the raytrace hider.

    And don't forget the speedup trick about making the hair opacity map truly b&w (on/off) and not grayscale, if it doesn't make the hair look too bad. It's very "oldschool", but transmapped hair IS oldschool.

  • Oso3DOso3D Posts: 15,045

    Turns out it wasn't the hair, it was the fact that UE2 bounce light with every skin grinds my render to a halt, particularly around the face.

    Trying to do realism without UE2, but... eh. Iray is so much easier. sigh

     

  • I posted a link to an alternative bounce light shader in the male render thread, did you try that?

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016

    OK, somebody the past few pages suggested using B&W only trans/opacity maps on hair, and I figured I'd give it a shot on a particular G3F style that is heavy handed with render times. Should be easy to do, go and find the opacity map, convert it to on/off, and give it a try. Problem, it's been a long day and my brain is very tiered. Looks like a lot of maps to convert and is not going to be as simple as I first thought it would be.

    OK, after getting over the shell shock of the number of opacity maps involved, I decide to look at the omni settings to see if there was something a tad simpler to try. The hair appears to only be using just Diffuse, Opacity, gloss1 and gloss2 at first glance. No velvet, no Subsurface, No translucence, No Refraction, (etc, etc, etc), Looks like a good candidate for just dropping the Daz Default on at first glance, till I scroll down to the technically overwhelming section.

    Now I'm completely lost. Is that sampling rate of 128 insane for a 90k-poly surface, or is that as light-footed as it can be?  (I'm going to make a pot of coffee and look at this some more, that dose not look right for some odd reason).

    WhatOpacityMapWasIt_001.png
    680 x 1200 - 394K
    Why_001.png
    380 x 440 - 13K
    Only_90k_Polys_001.png
    512 x 840 - 506K
    Post edited by ZarconDeeGrissom on
  • mjc1016mjc1016 Posts: 15,001
     

    Now I'm completely lost. Is that sampling rate of 128 insane for a 90k-poly surface, or is that as light-footed as it can be?  (I'm going to make a pot of coffee and look at this some more, that dose not look right for some odd reason).

    That's about as light as it gets...sampling rates of 1 are 'even'.  Lower than 1 are 'oversampling' and higher than 1 are 'undersampling'.  A rate of 0.001 would be totally, absolutely insane, besides being something in the neighborhood of 'centuries' for render times.

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016
    mjc1016 said:
     

    Now I'm completely lost. Is that sampling rate of 128 insane for a 90k-poly surface, or is that as light-footed as it can be?  (I'm going to make a pot of coffee and look at this some more, that dose not look right for some odd reason).

    That's about as light as it gets...sampling rates of 1 are 'even'.  Lower than 1 are 'oversampling' and higher than 1 are 'undersampling'.  A rate of 0.001 would be totally, absolutely insane, besides being something in the neighborhood of 'centuries' for render times.

    lol.  OK, was not sure, and just finished a quick test of NoHair, SR1 and SR128, that confirms as much.  It's nice hair and I want to use it, it's just that half-hour penalty for having it in a scene is quite painful.

    No Hair      3 minutes 35.51 seconds
    Hair  SR1    36 minutes 29.24 seconds
    Hair  SR128  30 minutes 29.51 seconds

    That at least gives me a base line to look at map possibilities. It dose look like the opacity maps are gray scale with only 256 unique colors, I'm just not sure yet how bad the hair will look with on/off maps or if that will effect any thing at all (yet)

    (EDIT) about half way threw the altered map tests, I realized I was fussing with the wrong opacity map for that hair. Looks like all the hair style maps from that PA are all in the same folder. That and the Opacity map info screen-cap was for the wrong one. Looking at the correct one, How can a 24bit jpg have that many unique shades of gray.

    (EDIT2) I think the PA already did some impressive tweaks, as I'm getting longer times with 4bit (16 color) and 1bit (2 color) B&W maps modified from the original jpg... (I used Irfanview to "Decrease color depth" and saved copies of the map to test.)

    1bit opacity map  34 minutes 46.62 seconds
    4bit opacity map  31 minutes 57.3 seconds

    1bit opacity map  1 days -31.-49 seconds ( *** Daz Studio midnight glitch)
    1bit opacity map  32 minutes 13.92 seconds
    4bit opacity map  30 minutes 29.3 seconds

    Yes, I reran the tests to be sure. very odd. I may have missed a simple step when saving the altered maps ("Check your staging", lol.  "Convert to gray scale" in Irfanview before saving.). I'll need to make the maps again after the current test that looks more promising.

    20161207_HairTest01_NoHair_001_Render 2.jpg
    600 x 600 - 177K
    20161207_HairTest01_SR1_001_Render 1.jpg
    600 x 600 - 181K
    20161207_HairTest01_SR128_Render 3.jpg
    600 x 600 - 181K
    OpacityMap_Stats2_001.png
    360 x 420 - 7K
    20161207_HairTest03c_map1bit_Render 9.jpg
    600 x 600 - 190K
    20161207_HairTest03c_map4bit_Render 10.jpg
    600 x 600 - 179K
    Post edited by ZarconDeeGrissom on
  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016

    OK, looking at that from a different angle for a brief side track.  A while back in Studio 4.7 I had done a few render tests on a floor with Opacity maps fir lights under the floor, and I found that maps that were on/off (1bit) had a drastically better render time. Tho as wowie had hinted at, simple test renders do not always convey the entire picture, and this hair is absolutely a good example of that. The original maps for the hair have almost no solid opaque areas at all, so decreasing color depth dose little (so far in testing), because the render engine must calculate each surface behind the semitransparent mesh, and on hair that adds up really fast.  So I decided to try altering the brightness and contrast of the map (In GIMP) to get more solid opaque areas to save some on rendering multiple layers of semitransparent meshes.

    After the former 1bit map tests I did not expect that to do much, Tho I gave it a shot. The first stab at enhancing the map gave an unexpected 8 unique colors (I didn't intend to go that far, oops).

    I opened that map, and decreased the color depth to 4bit and then 1bit consecutively for the final test maps.

    Between 1 and 4bit was no drastic difference in render time at all. However compared to the much more transparent original, drastic improvement in render times (It dose lack a bit of quality tho in appearance, I'll need to try that contrast thing in GIMP again).

    Enhanced Maps.
    1bit Enh opacity map  16 minutes 31.47 seconds
    4bit Enh opacity map  16 minutes 13.82 seconds
    24bit Enh opacity map  16 minutes 28.79 seconds

    I can almost live with that render time, if it did not look so bad at some viewing angles and in some spots.

    I do find it odd that the 1bit (on/off) map did not show similar advantages over values between 0% and 100% opaque as former tests had shown with just the opacity dial and no maps. It may just ba a factor of the amount of layers involved with hair, or that a map value somehow differs inside 3DL/Studio from just a dial setting somehow, I don't know.

    MapReContrast001.png
    440 x 150 - 17K
    Original_Vs_Enh1_OpacityMaps_001.png
    640 x 512 - 36K
    Enh1_1bit_Vs_4bit_OpacityMaps_001.png
    680 x 280 - 27K
    20161207_HairTest04a_mapE1bit_Render 12.jpg
    600 x 600 - 188K
    20161207_HairTest04a_mapE4bit_Render 11.jpg
    600 x 600 - 187K
    20161207_HairTest04a_mapE24bit_Render 13.jpg
    600 x 600 - 188K
    Post edited by ZarconDeeGrissom on
  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016

    Enough of that, I think I got it to a point I can live with.

    9 minutes 10.55 seconds

    Using the 'Levels' thing instead of the Brightness dials... I did discover that for some odd reason, 4bit gray scale Opacity maps do not go completely opaque giving the scalp a halo across the forehead. 24 bit gray scale and 1bit B&W maps work fine there tho.  Now that I figured out a path of a sorts, I have a few other hair styles to look at. I have a seeking suspicion that the more then 256 unique colors on the 24bit JPG files may be more color then gray scale, tho I don't think that is a major problem with the slow hair render times.

    20161207_HairTest05a_mapEnh5_24bit_Cam2_Render 16.jpg
    600 x 600 - 194K
    Gimp_Color_Levels_Enh5_001lbl1.png
    512 x 480 - 180K
    Post edited by ZarconDeeGrissom on
  • Zarcon, were you testing the maps with occlusion overridden like in the screencaps, or with occlusion off?

    ...I'd start with turning it off whatsoever, the hair is dark, it may be unnoticeable...

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016

    Ah, morning, I left the shader preset confusion settings all stock when I was fussing with the opacity maps (128 samples, atc. just like the settings screen cap of them). I don't like breaking settings that I don't know what they do (or don't remember what they are). What is occlusion  (exclusion/inclusion misspelling?)?... I need coffee.

    I think that was supposed to be dark-brown according to the preset, tho it dose look a tad more like black. Perhaps I clicked the wrong color, I don't remember (didn't matter for that opacity map testing, I should try a lighter color to be Shure). the race is on between loading that scene, and the coffee pot, lol.

    Yea, looks like the scalp needs a bit more opacity work on the lighter color presets. Looks OK on the darker ones.

    Enh5_Opas_Vs_Pink_Preset_001.png
    620 x 480 - 545K
    Enh5_Opas_Vs_XBlackRed_Preset_001.png
    620 x 440 - 453K
    Enh5_Opas_Vs_Blonde_Preset_001.png
    620 x 440 - 508K
    Enh5_Opas_Vs_Black_Preset_001.png
    620 x 440 - 438K
    Enh5_Opas_Vs_Brown_Preset_001.png
    620 x 440 - 459K
    Enh5_Opas_Vs_BrownDark_Preset_001.png
    620 x 440 - 419K
    Post edited by ZarconDeeGrissom on
  • "Occlusion" stands for "ambient occlusion" there. Basically it tells UE2 (the light that does AO) to use different sampling with "override" (you can set a surface to be shaded more precisely or more crudely as compared to the rest of the scene), or not to calculate AO for this surface at all, when the occlusion visibility is set to "no". 

  • mjc1016mjc1016 Posts: 15,001

    Zarcon, were you testing the maps with occlusion overridden like in the screencaps, or with occlusion off?

    ...I'd start with turning it off whatsoever, the hair is dark, it may be unnoticeable...

    Umm...yeah, for dark (anything darker than a medium brown and even some lighter colors) hair it's not really noticeable at all.

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016

    OK, thanks, and I apologies for the delay (there is a dance tonight and a GBFSC event tomorrow I had to do some perp for).  I'll do some render tests as soon as I can to see what effect that setting dose for that hair (Thanks for the clarification Kettu).

    Black Occlusion ON sr128  15 minutes 14.61 seconds
    Black Occlusion OFF       13 minutes 16.69 seconds
    Blonde Occlusion ON sr128 15 minutes 14.22 seconds
    Blonde Occlusion OFF      13 minutes 33.46 seconds
    Brown Occlusion ON sr128  15 minutes 42.1 seconds
    Brown Occlusion OFF       13 minutes 38.84 seconds

    A quick preliminary run dose not look as promising as fussing with the bit-depth range of the opacity maps. More testing is need tho... I think I see a patron forming, hmmm.

    20161209_HairTest01_OcluOff_Render 2.jpg
    600 x 600 - 190K
    20161209_HairTest01_OcluOn128_Render 1.jpg
    600 x 600 - 190K
    20161209_HairTest01blnd_OcluOn128_Render 3.jpg
    600 x 600 - 218K
    20161209_HairTest01brn_OcluOn128_Render 5.jpg
    600 x 600 - 200K
    20161209_HairTest01brn_OcluOff_Render 6.jpg
    600 x 600 - 200K
    20161209_HairTest01blnd_OcluOff_Render 4.jpg
    600 x 600 - 217K
    Post edited by ZarconDeeGrissom on
  • Zarcon, are those times with "progressive" on or off in the render settings?

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited December 2016

    Yea, there all progressive. It's a bit more painful to do spot renders in the other mode, because it can take that much longer for squares to fill in to what you wanted to see the adjustment of. That and I suspect the ratio of time it takes to render vs the different settings would not be that much different (about 47% faster with opacity map mods vs only 14% faster with just Occlusion off).  That and I don't have another twelve hours today to run every single one of the renders again with regular render mode, not including surface set up time.

    Post edited by ZarconDeeGrissom on
  • Progressive is the fastest for this sort of task, so no need to go all glacial with REYES =)

  • rofl.  After the in-progress render, I'm going to look into making actual soft-box geometry for my test chamber soft boxes. The exposed hanging in mid air Uber area light panel is kind of lacking for some things.

  • Shader mixer IDL camera vs UE2 bounce mode: 1 : 0

    I got a free scene up to pick apart: http://www.daz3d.com/forums/discussion/comment/1892266/#Comment_1892266

Sign In or Register to comment.