3Delight Laboratory Thread: tips, questions, experiments

11920222425100

Comments

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969


    First with a sphere that is making shadows on G2F. No GI. The part in shadow gets some specular somehow (top of the head for ex)


    That's the Edge cheat. Kill it. It won't be on by default.


    I spot the problem

    In fact that is not the 100% SSS which causes the problem. It seems you have something wrong in the ambient

    Third render. I saw that G2F Top and short get colored in shadow and no specular. And I thought that must be the ambient channel. So I put 0% ambient on the top and bingo. It gets black.

    I didn't have a look at your code. There is no control of the ambient channel on your shader so there is no way to control that but there is something

    Okay I am stupid. I don't understand what you mean. Again, could you please draw arrows on your images?


  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969


    Imagine, it's the night or you are in a cave with just a lamp torch. What is the environment color? Black. If you put a black map into your GI light it's the same effect as not putting any map. Try it

    If you are primarily interested in nights and caves, then sure =) If it's daylight and you're on a planet with an atmosphere in a room that opens to the skies via windows (or lack of walls), then you need an environment.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Other problem : the default glass transmission hitmode parameter and shadow color being inverted or something like that

    I'm not on the computer, but yes the shader shadow colour should be inverted because it's a coloured shadow thing for artists. If you put green on there, you will get green shadow = transmission Oi is purple-pink, 1-green.

    The primitive hitmode shadow is controlled via basic shadow opacity slider. They are not related.

    You only need shader hitmode when you don't use photon-mapped caustics. It works for fake caustics.

    It's not too intuitive, but the docs will explain it.

    There is a more noticeable bad default in there, white photon opacity. I'll put a clamp in there and change the default. For now, just use a meaningful shade of gray (or colour!) for photon mapped caustics.

    I thought of linking photon opacity with absorption colour, but decided against it for reasons of more artistic control.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Rogerbee said:

    I tried out a V4 map and that glowing nostril effect still persists, even when you turn off that edge setting completely. Something else is awry and I'm not sure what. The texture is Farissa by Virtual World

    Unfortunately I don't have the V4 UV for G2F. Probably don't have this particular texture, either.

    This looks a 'too much diffuse light energy' kinda glow. As if there is a lighter-than-skin colour in the nostrils. May be a UV issue or a particular texture problem.

    What do you get with V4 default textures? I think the V4 hero pack is currently free, it has the real V4 and her Pro bundle with hi-res textures. I've had it for quite a while, we could compare results on my V4 and your G2F.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited June 2015

    mjc1016 said:
    I'm going to paste those Oren-Nayar values here...there's more of you want to dig through the data from the O-N database...

    I still haven't figured out how to use their DB.

    Is there anywhere an answer to Wowie's question? Half the PDF's never load for me.

    Rogerbee, wherever you see 'diffuse roughness' in my shaders. But of course, it's all a starting point for artistic choices.

    Post edited by Mustakettu85 on
  • mjc1016mjc1016 Posts: 15,001
    edited December 1969

    mjc1016 said:
    I'm going to paste those Oren-Nayar values here...there's more of you want to dig through the data from the O-N database...

    I still haven't figured out how to use their DB.

    Is there anywhere an answer to Wowie's question? Half the PDF's never load for me.

    Rogerbee, wherever you see 'diffuse roughness' in my shaders. But of course, it's all a starting point for artistic choices.

    I'll answer with a "I'm not sure", because I haven't waded through all the ones I can get to load...yet. But the few tests I've done with those values seem to work well with whatever the 'standard' Oren-Nayar is.

    A good example would be the one for skin. I'd imagine that a baby would have a somewhat lower roughness value than a 60 year old farmer...

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


    First with a sphere that is making shadows on G2F. No GI. The part in shadow gets some specular somehow (top of the head for ex)


    That's the Edge cheat. Kill it. It won't be on by default.


    I spot the problem

    In fact that is not the 100% SSS which causes the problem. It seems you have something wrong in the ambient

    Third render. I saw that G2F Top and short get colored in shadow and no specular. And I thought that must be the ambient channel. So I put 0% ambient on the top and bingo. It gets black.

    I didn't have a look at your code. There is no control of the ambient channel on your shader so there is no way to control that but there is something

    Okay I am stupid. I don't understand what you mean. Again, could you please draw arrows on your images?


    That could be your edge cheat. If you put every paarameter at 0 and every color at black you still get some shading instead of a complete black render. However I'm really not sure. I made another test render in very difficult lightning : two G2F inside a box on which I made a hole in it and inverted normals. Added a physDistant light and a delightGI with no env map. One of the G2F has your preset (left), the one on the right is the base one. The two figures will mostly get IDL from the reflected light going through the hole. It's not only the edge, but the whole surface which seem to be mixed with a white color


    Imagine, it's the night or you are in a cave with just a lamp torch. What is the environment color? Black. If you put a black map into your GI light it's the same effect as not putting any map. Try it

    If you are primarily interested in nights and caves, then sure =) If it's daylight and you're on a planet with an atmosphere in a room that opens to the skies via windows (or lack of walls), then you need an environment.

    Not really but there are cases where you don't need any, especially in dark scenes

    Other problem : the default glass transmission hitmode parameter and shadow color being inverted or something like that

    I'm not on the computer, but yes the shader shadow colour should be inverted because it's a coloured shadow thing for artists. If you put green on there, you will get green shadow = transmission Oi is purple-pink, 1-green.

    The primitive hitmode shadow is controlled via basic shadow opacity slider. They are not related.

    You only need shader hitmode when you don't use photon-mapped caustics. It works for fake caustics.

    It's not too intuitive, but the docs will explain it.

    There is a more noticeable bad default in there, white photon opacity. I'll put a clamp in there and change the default. For now, just use a meaningful shade of gray (or colour!) for photon mapped caustics.

    I thought of linking photon opacity with absorption colour, but decided against it for reasons of more artistic control.

    Forget this one. It's my head not working after too much boring test renders

    INSIDEABOX.jpg
    1358 x 848 - 482K
    G2F_DELTA_spot01.jpg
    699 x 846 - 56K
  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    wowie said:

    It's not specific to the shader.

    DS have improved HD morphs support for quite some time now. Although it's slightly slower when rendering (depending on what subD level the morphs were built in), there's no more long load times with HD morphs.

    I didn't mention the shader actually. It's probably the render script and how it makes 3Delight behave. Set normally, 3Delight sits while it pre-calculates what it's doing with the HD. Kettu's script seems to speed this up, which makes HD a much better proposition.

    CHEERS!

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969


    Unfortunately I don't have the V4 UV for G2F. Probably don't have this particular texture, either.

    This looks a 'too much diffuse light energy' kinda glow. As if there is a lighter-than-skin colour in the nostrils. May be a UV issue or a particular texture problem.

    What do you get with V4 default textures? I think the V4 hero pack is currently free, it has the real V4 and her Pro bundle with hi-res textures. I've had it for quite a while, we could compare results on my V4 and your G2F.

    It isn't the textures as there was the same issue with V6 and it's not the lighting as I used the same lighting with every figure. There's something in the way you have set up the shader for V6 or G2F. You don't get the effect with G2M/M6. I'll have a look at the textures with the lights I used for Nevio and Boris and see what I get there.

    CHEERS!

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    mjc1016 said:
    I'm going to paste those Oren-Nayar values here...there's more of you want to dig through the data from the O-N database...

    I still haven't figured out how to use their DB.

    Is there anywhere an answer to Wowie's question? Half the PDF's never load for me.

    Rogerbee, wherever you see 'diffuse roughness' in my shaders. But of course, it's all a starting point for artistic choices.

    Ah, gotcha, I'll give one a whirl sometime.

    CHEERS!

  • wowiewowie Posts: 2,029
    edited December 1969

    Rogerbee said:

    I didn't mention the shader actually. It's probably the render script and how it makes 3Delight behave. Set normally, 3Delight sits while it pre-calculates what it's doing with the HD. Kettu's script seems to speed this up, which makes HD a much better proposition.

    Sorry, no. It's not the render script.

    I've been rendering HD details with IPR and they all finish about the same as non HD version.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Rogerbee said:

    It isn't the textures as there was the same issue with V6 and it's not the lighting as I used the same lighting with every figure. There's something in the way you have set up the shader for V6 or G2F. You don't get the effect with G2M/M6.

    Scientific method suggests you apply a G2M preset to G2F and see what happens.

  • wowiewowie Posts: 2,029
    edited December 1969


    Scientific method suggests you apply a G2M preset to G2F and see what happens.

    And the G2F preset to G2M. :)

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969


    That could be your edge cheat. If you put every paarameter at 0 and every color at black you still get some shading instead of a complete black render. However I'm really not sure. I made another test render in very difficult lightning : two G2F inside a box on which I made a hole in it and inverted normals. Added a physDistant light and a delightGI with no env map. One of the G2F has your preset (left), the one on the right is the base one. The two figures will mostly get IDL from the reflected light going through the hole. It's not only the edge, but the whole surface which seem to be mixed with a white color

    The circles you marked are the edge cheat. Does the one in the box have edge off? Edge is actually a very high roughness ancient 3Delight specular model (moreover, it's the L-vector accepting version which is stuck in an illuminance loop... either I have overlooked a check there, or it's buggy internally). It's driven by fresnel, but it can light up the surface overall (very slightly in 'normal' light).

    Diffuse ray hits are dealt with explicitly after a rayinfo() check, no specular components are supposed to be picked up.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    wowie said:

    Scientific method suggests you apply a G2M preset to G2F and see what happens.

    And the G2F preset to G2M. :)

    Exactly! :)

    As for HD: IPR and my scripts all call the raytracer. What's the REYES performance like with them?

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    Rogerbee said:

    It isn't the textures as there was the same issue with V6 and it's not the lighting as I used the same lighting with every figure. There's something in the way you have set up the shader for V6 or G2F. You don't get the effect with G2M/M6.

    Scientific method suggests you apply a G2M preset to G2F and see what happens.

    You'd get a mess as they are not the same UV's, this is G2 and not Genesis.

    CHEERS!

  • RogerbeeRogerbee Posts: 4,460
    edited June 2015

    Exactly! :)

    As for HD: IPR and my scripts all call the raytracer. What's the REYES performance like with them?

    I was merely going by how HD renderered with everything at defaults as I did it before I knew of the 3Delight thread. Maybe it's improved some generally. Whatever, it has prompted me to get the HD add on for Darius on Wednesday.

    CHEERS!

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

    Rogerbee said:

    Exactly! :)

    As for HD: IPR and my scripts all call the raytracer. What's the REYES performance like with them?

    I was merely going by how HD renderered with everything at defaults as I did it before I knew of the 3Delight thread. Maybe it's improved some generally. Whatever, it has prompted me to get the HD add on for Darius on Wednesday.

    CHEERS!

    You may have this impression because you don't have to wait for 3delight to prepare the mesh for subsurface which was the case with the old SSS implementation. Now with RTSS there is no wait time before the shooting of rays

  • mjc1016mjc1016 Posts: 15,001
    edited December 1969

    Rogerbee said:

    Exactly! :)

    As for HD: IPR and my scripts all call the raytracer. What's the REYES performance like with them?

    I was merely going by how HD renderered with everything at defaults as I did it before I knew of the 3Delight thread. Maybe it's improved some generally. Whatever, it has prompted me to get the HD add on for Darius on Wednesday.

    CHEERS!

    You may have this impression because you don't have to wait for 3delight to prepare the mesh for subsurface which was the case with the old SSS implementation. Now with RTSS there is no wait time before the shooting of rays

    And depending on the SSS, mesh and other settings...that could be a considerable amount of time.

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    I'm just glad it all works!

    CHEERS!

  • RogerbeeRogerbee Posts: 4,460
    edited June 2015

    A bit more HD now. Nevio with a lot of dial tweaking. Even with SubD set to 2 it doesn't increase the render time. I also set the ON diffuse roughness to the value MJC suggested.

    CHEERS!

    Nevio_Defined_RTK_02.jpg
    577 x 750 - 269K
    Nevio_Defined_RTK_01.jpg
    577 x 750 - 310K
    Post edited by Rogerbee on
  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    LOL! I didn't turn up the SubD when I did those Borius renders. In his video for Boris, Smay turns up the SubD to 3, so I did too and still with no increase in the render time. Ordinarily it's the SubD that makes 3Delight stumble when it begins the render, but not with RTK. This is a big plus and one worth mentioning in the release notes.

    CHEERS!

    Borius_HD_SubD_3.jpg
    577 x 750 - 272K
  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    Daz have done a very cool thing, when you adjust the viewport SubD, they've linked that slider to the render SubD so you adjust both at once. It would seem that 3 is the best for HD and supposedly the most my 16gb of RAM can handle.

    Here's Nevio again at SubD 3

    CHEERS!

    Nevio_Defined_SubD_3.jpg
    577 x 750 - 269K
  • wowiewowie Posts: 2,029
    edited June 2015

    mjc1016 said:

    And depending on the SSS, mesh and other settings...that could be a considerable amount of time.

    Did a quick test with a single G2F/V6, with and without the HD morphs. With progressive rendering.

    With US2, but without HD morphs - 1 min 24.27 secs
    With US2, HD morphs and subD at level 3 - 3 min 19.68 secs
    With US2, HD morphs and subD at level 2 (default) - 3 min 21.54 secs
    With Radium, HD morphs and subD at level 2 (default) - 4 min 24.26 secs


    You may have this impression because you don't have to wait for 3delight to prepare the mesh for subsurface which was the case with the old SSS implementation. Now with RTSS there is no wait time before the shooting of rays

    Don't know about others, but from those test results, Takeo is correct. It's just an impression since users didn't see a pause (no precompute pass). But in the end, it still ended up being slower. Of course, this is with a single figure. Based on experience, a single G2F/G2M precompute pass can take up to 20 sec (on my machine) per figure. So I would say, pass more than 3 figures, the raytraced SSS will be faster overall.

    But on single figures and close up shots, it will be slower.

    Post edited by wowie on
  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    Try it without progressive rendering. I have it turned off so I can adjust the pixel filtering.

    CHEERS!

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Rogerbee said:

    Try it without progressive rendering. I have it turned off so I can adjust the pixel filtering.

    I hope you aren't rendering my shader in a default REYES mode, because outside of the raytracer this must take forever!

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    wowie said:

    With US2, but without HD morphs - 1 min 24.27 secs
    With US2, HD morphs and subD at level 3 - 3 min 19.68 secs
    With US2, HD morphs and subD at level 2 (default) - 3 min 21.54 secs

    IIRC, 3Delight always renders the limit surface for SubD (today, not the old "DS3" type of SubD). So when you go HD, there is a jump, but other than that SubD levels don´t matter anymore. The less-than-one-second difference is within the margin of error, I'd say.

    Limit surface is explained here: http://en.wikipedia.org/wiki/Subdivision_surface

  • RogerbeeRogerbee Posts: 4,460
    edited June 2015

    Rogerbee said:

    Try it without progressive rendering. I have it turned off so I can adjust the pixel filtering.

    I hope you aren't rendering my shader in a default REYES mode, because outside of the raytracer this must take forever!

    I'm using your Raytracer Extra HQ script, always have since I got RTK. The renderer option for Progressive is at 0 which I assume is the equivalent to turning it off in ordinary 3Delight.

    Wowie was doing all his tests with progressive on which wouldn't be comparable to RTK with the script I'm using.

    Here's another fave texture of mine, the M5 Dave with M6HD turned on, the SubD at 3 and the diffuse roughness at MJC's setting. The spec maps you did are fine for him as the G2M UV is, I believe, the same as M5.

    CHEERS!

    PS (Don't pay too much attention to the background. I had to turn 'cast shadows' off on the building as it was casting a shadow on the cyclorama behind it making it look like a backdrop.)

    Dave_HD_RTK.jpg
    577 x 750 - 342K
    Post edited by Rogerbee on
  • mjc1016mjc1016 Posts: 15,001
    edited December 1969

    Also the precompute pass can take a very long time...with certain combinations of of lights along with the other SSS shaders. AoA's is much worse than US/US2, especially with GI lighting.

  • wowiewowie Posts: 2,029
    edited June 2015


    IIRC, 3Delight always renders the limit surface for SubD (today, not the old "DS3" type of SubD). So when you go HD, there is a jump, but other than that SubD levels don´t matter anymore. The less-than-one-second difference is within the margin of error, I'd say.

    Limit surface is explained here: http://en.wikipedia.org/wiki/Subdivision_surface

    Oh, I know that. HD just means there's more geometry data to be loaded. I generally don't use them, particularly since most HD morphs don't really look that good. I still prefer using displacement for details.

    I think most of the HD support improvements were made in the 4.7 builds.


    Try it without progressive rendering. I have it turned off so I can adjust the pixel filtering.

    Sorry. No even an option. Scripted 3delight doesn't work with IPR. Pixel filtering is just a hack anyway. If you want the highest quality, render at very high res. Twice the pixel density should be the goal, so if you want 1920x1080, render at 3840x2160.

    Post edited by wowie on
Sign In or Register to comment.