3Delight Laboratory Thread: tips, questions, experiments

11516182021100

Comments

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

    mjc1016 said:

    I don't think what you're going to do with your tuple is physically correct but I'm sure some people will be happy with it

    It's an approximation, granted; but it's not breaking rules, just being kinda 7th grade physics when all the formulas are algebra, not college-level when you move on to more correct partial derivatives LOL

    Look at the pic of the prism on the right: http://en.wikipedia.org/wiki/Dispersion_(optics)

    The real spectrum is continuous; using a 3-tuple for eta will mean I'm dividing it roughly into three parts. It's like with numerical modeling in continuum mechanics (deforming structures etc): you can have a mesh with a resolution of 5 cells per unit of length, or a mesh with 50 cell/unit, etc. The first will be crude, but it will give you a general idea of the process and it may even be enough (depending on scale, complexity of the process etc.).

    It should be possible to write a dedicated procedure for finer division, but I don't really want to go that deep in foreseeable future (I'm hoping that we'll get native spectral colour support one day).


    The frequencies typically used are the RGB red, green, blue...and I have one, but it's old and slow.

    The easiest 'advance' is to use 6 frequencies. You know the typical 'rainbow'...

    I won't try to discuss math because I don't remember all the details but I know I thought about making a dll for some calculations that couldn't be done just with RSL at that time. And I also didn't count on the photon mapping shadeop because there was something that wasn't right in it for the purpose
    @mjc1016 I think you're trying to sell me a picture of a fake method. All old pixar or 3delight shaders I've seen used some fake trick. If that is the real thing then it should project some colored light if you put a plane somewhere

    My goal when I studied the thing was to do some beautiful renders like http://zeealpal.deviantart.com/art/Simple-is-Beautiful-Caustics-263606354.

    Pour liquids onto/into static meshes, spray vapours into the "air", simulate sand, flames etc. I think Blender can help me with liquids, if I export the result as OBJ, but I don't know about stuff like sand or vapours (they're not "OBJ-able", are they?).

    In theory, I should even be able to write something like this myself (the physics of it all, which is continuum mechanics, is what I deal with daily), but as I am a "mad scientist" (TM) and not an IT specialist, I never had any training in optimising the code to make it fast, so whatever I write will most likely take forever to solve anything but the crudest resolutions.

    Smoke sim? http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.65/Smoke_Simulation

    I don't really see your goal. May be some scientific simulation tools with python coding would be more appropriate

    The answer's in the docs actually, the part that is already up ;)
    Because it's not supported - the API for that class changed for 4.7, and the scripting docs couldn´t keep up with the changes. I have yet to check if the changes are documented now...

    http://www.daz3d.com/forums/discussion/49075/

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

    Rogerbee said:

    I tried one of Darius,

    Now, here's an interesting thing, I took Bjorn, dialled out M6 and put in Darius instead. Then I loaded his textures into the appropriate bits of the shader. I forgot to change the UV from Michael 6, but, the textures worked. When I tried to change the UV to Darius6 there was a big gap under his chin. Odd, so, it's possible that the 'custom UV' for Darius6 is really M6, but why would the texture be off if I try loading that UV!?

    Oh well, it works, as you will now see...

    CHEERS!

    PS (The lights are the Delight GHDRI and some from the Fantasy Lights set.)

    Seems nice. What about more complex scene and lights?

  • wowiewowie Posts: 2,029
    edited December 1969

    Rogerbee said:

    It doesn't look like a DS render at all to me, and that is what is so brilliant about it



    I tried one of Darius,

    Still looks too transluscent, too much SSS.

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969


    Seems nice. What about more complex scene and lights?

    It's on the to do list....

    CHEERS!

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    wowie said:
    Rogerbee said:

    It doesn't look like a DS render at all to me, and that is what is so brilliant about it



    I tried one of Darius,

    Still looks too transluscent, too much SSS.

    Take that up with Kettu, not me, I didn't adjust anything she did.

    CHEERS!

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Smoke sim? http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.65/Smoke_Simulation

    I don't really see your goal. May be some scientific simulation tools with python coding would be more appropriate

    Is there a way to get the results of that smoke sim into DS?

    I think the goal is obvious, Takeo. I want an integrated solution for simulating processes that are way too hard to model by hand, but the results of which are important for visual storytelling.

    Thank you for the Python idea; my brother is a huge fan of this language, I will ask him if he knows of some good physics lib and of suitable exporters/importers.


    http://www.daz3d.com/forums/discussion/49075/

    Thanks, but Rob says there that the API will change... have you made sure the sample he posted reflects the current state of affairs?

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Rogerbee said:

    Take that up with Kettu, not me, I didn't adjust anything she did.

    Now is a good time to learn how to use the included CopyMaps scripts: select the skin surfaces (Default Templates 1, 2 and 3 in the Surfaces tab selection list) and run the script that copies SSS maps to diffuse. Then dial in some diffuse strength. If you use the mix mode instead of adding, the only strength you need to adjust is the diffuse one; it will automatically balance the SSS contribution. Adding means you will need to decrease SSS strength manually (more artistic control).

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    mjc1016 said:

    The frequencies typically used are the RGB red, green, blue...and I have one, but it's old and slow.

    The easiest 'advance' is to use 6 frequencies. You know the typical 'rainbow'...

    Yup, exactly. The 3-tuple eta supported for trace() now is an array of IoRs for R, G and B. It may even be enough for us, let's wait and see what I manage to get.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Rogerbee said:

    As I replied later it doesn't matter about the backdrop, I'll simply replace it with a scene item. Glad you like Darius, I loved that all I had to change was the textures. I think people are really going to like this shader as you can get great results without any fiddling. The shader settings are straight out of the box, no tweaks needed. All I did do was to change the pixel filter to Catmull Rom. I'll try a female figure next.

    And the habit of using a scene item for backdrop will also mean that when you add volumetrics, they will always have geometry to actually render against.

    How are your render times now that you've got the scripts working? And there is a mini-tutorial in the docs that explains how to make a frontend script with settings you prefer. Although I guess now that 4.8 saves even the scripted rendering settings per scene (which I love!!), you can just make a new default scene.

    The trickiest part to get right is the eyes - you may have noticed these settings change the most between characters, and I made control maps to counter the fact that G/G2 eye geometry isn't optimal for RT SSS, the iris not being continuous geometry with the sclera.

    The eyes also show using the Overlay channel to add just the veins on top the SSS.

    I've started building my own universal eye replacer models that should (I hope) work better and easier with my shaders. This will take some time, of course, before they are release-ready.

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

    Is there a way to get the results of that smoke sim into DS?

    I think the goal is obvious, Takeo. I want an integrated solution for simulating processes that are way too hard to model by hand, but the results of which are important for visual storytelling.

    Thank you for the Python idea; my brother is a huge fan of this language, I will ask him if he knows of some good physics lib and of suitable exporters/importers.


    No idea. Never tried.

    Thanks, but Rob says there that the API will change... have you made sure the sample he posted reflects the current state of affairs?

    Didn't try either!; I barely use scripted renders

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    Rogerbee said:

    Take that up with Kettu, not me, I didn't adjust anything she did.

    Now is a good time to learn how to use the included CopyMaps scripts: select the skin surfaces (Default Templates 1, 2 and 3 in the Surfaces tab selection list) and run the script that copies SSS maps to diffuse. Then dial in some diffuse strength. If you use the mix mode instead of adding, the only strength you need to adjust is the diffuse one; it will automatically balance the SSS contribution. Adding means you will need to decrease SSS strength manually (more artistic control).

    Ok will do, mind you I don't think anything needs adjusting.

    Here's Anna in the meantime...

    CHEERS!

    Anna_RTK.jpg
    577 x 750 - 213K
  • RogerbeeRogerbee Posts: 4,460
    edited May 2015

    Rogerbee said:

    As I replied later it doesn't matter about the backdrop, I'll simply replace it with a scene item. Glad you like Darius, I loved that all I had to change was the textures. I think people are really going to like this shader as you can get great results without any fiddling. The shader settings are straight out of the box, no tweaks needed. All I did do was to change the pixel filter to Catmull Rom. I'll try a female figure next.

    And the habit of using a scene item for backdrop will also mean that when you add volumetrics, they will always have geometry to actually render against.

    How are your render times now that you've got the scripts working? And there is a mini-tutorial in the docs that explains how to make a frontend script with settings you prefer. Although I guess now that 4.8 saves even the scripted rendering settings per scene (which I love!!), you can just make a new default scene.

    The trickiest part to get right is the eyes - you may have noticed these settings change the most between characters, and I made control maps to counter the fact that G/G2 eye geometry isn't optimal for RT SSS, the iris not being continuous geometry with the sclera.

    The eyes also show using the Overlay channel to add just the veins on top the SSS.

    I've started building my own universal eye replacer models that should (I hope) work better and easier with my shaders. This will take some time, of course, before they are release-ready.

    The render times aren't too bad at all, well, when you're not rendering hair that is. Anna took more than 20 mins. Regarding eyes, when you use that Iris morph, what is the optimum setting, I've just been dialling it to 100%. I'm sure any work you do will be beneficial.

    CHEERS!

    Post edited by Rogerbee on
  • wowiewowie Posts: 2,029
    edited December 1969

    Rogerbee said:

    Here's Anna in the meantime...

    Still the same problem - too translucent. Very wax looking.

    Plus there's some backlight issues on the nostrils and lips.

    If you want to tweak SSS, it's best to not use maps at all. Here's a render with and without diffuse maps.

    SSS can generally be seen everywhere on the skin, but backscatter is limited to just certain parts.

    2.jpg
    823 x 1070 - 191K
    1.jpg
    823 x 1070 - 235K
  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    Yet again, I didn't do anything with those settings, they are as Kettu set them up. You're telling the wrong person!

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    I didn't scroll down the shader when I switched maps from Bjorn to Darius. There were other maps to change and then the UV thing made sense as well. I then tried Neo too. I am happy with the results so far. Bear in mind that this is an Alpha of RT Kit, so if there is anything that you guys think needs adjusting, suggest it to Kettu and she can maybe adjust it for the full release. She knows better than I do what the settings do. I'll study them in more depth when the full version comes out.

    CHEERS!

    Neo_RTK.jpg
    577 x 750 - 175K
    Darius_RTK.jpg
    577 x 750 - 176K
  • wowiewowie Posts: 2,029
    edited May 2015

    Rogerbee said:

    Yet again, I didn't do anything with those settings, they are as Kettu set them up. You're telling the wrong person!

    If you noticed a few pages back, I did tell that to Kettu about her Jayden render and she answered it was a stylistic decision.

    Those are your renders, so ultimately, it's you the one that sets them up. Not Kettu.

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

    I load the textures and pose the figure, I didn't adjust any of the shader settings, those were done by Kettu, I didn't change them. They are as they were set up by her, so, if you think those settings need changing, talk to her about them. If she takes on board what you say and adjusts them prior to the release then so be it. I am just testing what she has done and that is all.

    CHEERS!

  • wowiewowie Posts: 2,029
    edited December 1969

    Rogerbee said:

    I load the textures and pose the figure, I didn't adjust any of the shader settings, those were done by Kettu, I didn't change them. They are as they were set up by her, so, if you think those settings need changing, talk to her about them. If she takes on board what you say and adjusts them prior to the release then so be it. I am just testing what she has done and that is all.

    The point of bug testing, be it through alpha, beta or release candidate versions is to detect problems. That means not just using it out of the box, but trying out all the features to see whether they work as intended.

    Then there's documentation, like what the features are, what they do, how they're structured etc. Are there specific cases where they should be enabled/disabled? Some example usage scenarios would be nice.

    Testing inputs are invaluable to developers. Plainly loading, applying and using it straight out only let them know it works (for you), which they know already.

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    I will get to that part when I am good and ready. Right now, I am seeing what different textures look like with that shader applied. Once I know which textures will work well with the shader, I will then look at the documentation and see what can be adjusted, but, I will also rely on her input for that as she knows the settings. Also, Kettu hasn't done MAT files for all the main textures yet. If there are ones that work well, then she can, if she has them, write MAT files for them. So, what I've been doing so far isn't entirely unnecessary.

    CHEERS!

  • MistaraMistara Posts: 38,675
    edited December 1969

    mjc1016 said:

    The frequencies typically used are the RGB red, green, blue...and I have one, but it's old and slow.

    The easiest 'advance' is to use 6 frequencies. You know the typical 'rainbow'...

    ROYGBIV :lol:

  • RogerbeeRogerbee Posts: 4,460
    edited May 2015

    My English teacher taught me a great way of remembering that: "Rotten old Yogi gave Boo-boo indigestion violently!"

    CHEERS!

    Post edited by Rogerbee on
  • MistaraMistara Posts: 38,675
    edited December 1969

    Every good boy does fine :)

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    This is a test to show what the M4 Rob texture looks like with RT Kit. There's clearly something not quite right there, Kettu, any ideas what it might be? This is the first I've tried with M4 maps.

    CHEERS!

    Rob_RTK_Odd.jpg
    577 x 750 - 195K
  • mjc1016mjc1016 Posts: 15,001
    edited December 1969

    Rogerbee said:

    This is a test to show what the M4 Rob texture looks like with RT Kit. There's clearly something not quite right there, Kettu, any ideas what it might be? This is the first I've tried with M4 maps.

    CHEERS!

    Talking about the 'crusty'/'flaky' appearance of his lips/side of face or the overall, dry, dusty appearance?

    The 'flaky' stuff is a combo of the angle and intensity of the rim light.

    As is, the rest of skin just seems lacking in specular response...which could be the control maps aren't right. They probably are getting 'corrected' as if they were color images, not greyscale or they are too light or dark (more likely) to begin with.

    As is, put a white skull face overlay on the face....and say you are doing a Baron Samedi render...

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Rogerbee said:

    Here's Anna in the meantime...



    Plus there's some backlight issues on the nostrils and lips.
    SSS can generally be seen everywhere on the skin, but backscatter is limited to just certain parts.

    Doesn't look like an SSS issue to me, but rather a problem with the backlight... It looks as if it's not casting a shadow (look at the top of her head that is somehow visibly shining through the hair!).

    Rogerbee, would you please check that scene? Each and every directional light should be casting shadows. Specular-only lights should only be used paired with diffuse-only lights.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited May 2015

    Rogerbee said:

    This is a test to show what the M4 Rob texture looks like with RT Kit. There's clearly something not quite right there, Kettu, any ideas what it might be? This is the first I've tried with M4 maps.

    What maps are you using to drive specular strength and/or roughness? As Mjc says, this is clearly the root.

    It looks to me that roughness might be too low to get cleanly resolved with this level of bump and pixel samples.

    I include roughness maps for G2M and G2F UVs, they also work acceptable for V6 and M6. If you can, try converting them for M4 UV and see if it helps. I'm sorry I don't have this texture set, so I don't know what the specular maps are like.

    Low roughness is polished material. If your map has values close to black, and you feed it to roughness, it's guaranteed to firefly like this. You could overlay an additive grey layer in LIE, that should help.

    And of course, when you change specular maps, be it strength or roughness, the multipliers will need adjustment because these maps are never done to some standard in DS world, they are all different.

    Post edited by Mustakettu85 on
  • Mustakettu85Mustakettu85 Posts: 2,933
    edited May 2015

    Rogerbee said:

    Also, Kettu hasn't done MAT files for all the main textures yet. If there are ones that work well, then she can, if she has them, write MAT files for them.

    Out of the primary G2 characters, I have M6, V6, and Teen Jayden. These are all done, along with Dawn and Dusk, it's just that the "rival" MATs and Jayden+Lance aren't packaged in the alpha.

    As I told Wowie, these presets reflect my own preference first, and then, they serve an educational purpose: they provide a "naked" SSS base for the user to learn the CopyMaps approach, via using the script to copy SSS maps to diffuse and then tweaking diffuse parameters to taste. Surfaces tab - Default Templates 1, 2 and 3 - Copy SSS Maps to Diffuse. This is the method. Getting used to selecting surfaces via those templates I believe to be an essential skill for any serious G2 user.

    Then, when the user progresses to updating other character sets that predate the kit, s/he will already have a foundation in using the scripts, and will be able to copy the bump into specular strength, for instance (because it often makes the most sense).

    Roughness is best driven by smooth, gradient maps outlining the general difference in reflectivity across the human body. Like my rather crude example ones provided, which are licenced as merchant resource, so you can convert them to any UV mapping, alter in any image editor, etc.

    You may find you don't even need a strength map when you have a good bump/displacement and a roughness map.

    Basically, I am a wannabe TD, as I always say. I write techie stuff, and then I can teach others to be their own lookdevs - adjusting the shaders AND maps to get the best look (which will always be different from project to project). I don't really make click-and-render presets, it's too boring for my principles. =)

    Post edited by Mustakettu85 on
  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    wowie said:

    Then there's documentation, like what the features are, what they do, how they're structured etc. Are there specific cases where they should be enabled/disabled? Some example usage scenarios would be nice.

    Testing inputs are invaluable to developers.

    You are absolutely right. Thank you.

    Example usage scenarios, like what I call "tips" generally? For instance, there will be a tip about Edge Strength being a cheat for "peach fuzz" that is mostly useful for closeups and only on specific surfaces.


    I will then look at the documentation

    I would be much obliged if you started reading the documentation right away. Granted, the shaders per se are not yet detailed there (however, with each update of the ODT file in the google folder, there are more and more things up there - I just uploaded a new iteration BTW), and there are WIP chapters like the one on using volumetric shaders on primitives, but the section on render scripts is pretty much finished.

    It would really help me if you actually used the documentation and then maybe suggested ideas about explaining something that I may have overlooked.

  • Mustakettu85Mustakettu85 Posts: 2,933
    edited December 1969

    Is there a way to get the results of that smoke sim into DS?

    No idea. Never tried.

    Thanks, but Rob says there that the API will change... have you made sure the sample he posted reflects the current state of affairs?

    Didn't try either!; I barely use scripted renders

    I'm thinking there may be a way to have the smoke in a RIB. Maybe. I'll try that 3Delight exporter for Blender ( http://mattebb.com/3delightblender/ ) and see, but not too soon.

    And I just checked the DS4 scripting page in the wiki here. I couldn't see any info about camera classes or background-related classes, and, well, the object list keeps growing, but very gradually. That's okay with me; I can wait.

  • RogerbeeRogerbee Posts: 4,460
    edited December 1969

    Rogerbee said:

    Here's Anna in the meantime...



    Plus there's some backlight issues on the nostrils and lips.
    SSS can generally be seen everywhere on the skin, but backscatter is limited to just certain parts.

    Doesn't look like an SSS issue to me, but rather a problem with the backlight... It looks as if it's not casting a shadow (look at the top of her head that is somehow visibly shining through the hair!).

    Rogerbee, would you please check that scene? Each and every directional light should be casting shadows. Specular-only lights should only be used paired with diffuse-only lights.

    Yeah, I can do that. The backlight used is Mid-Top from your Fantasy Lights set and it is set to cast shadows. The intensity is at 100%, I can reduce that and see what happens.

    CHEERS!

Sign In or Register to comment.