[jfrancis]

Hello forum, its been a while since I posted here... (my pc was on-site for a couple of months, for a project)

Couple of 3.2 boolean tests;

red glass (priority 2), clear glass swirls (priority 1), swisscheese air holes (priority 0)



and one with white sss swirls, I'm seeing that roughness 0 messes up these materials, they lose the attenuation (issue is not related to booleans)



Using an 'air' material to boolean is pretty neat, but the material will block diffuse light (just like any transmissive can come up against the caustics of caustics issue). i.e. I tried to boolean holes out of the white sss material but the holes turned dark grey. Obviously it works with glass as in the first render, because the glass appearance wont be affected by the lack of GI.

Interesting

[eric nixon]

One more 3.2 test, nested booleans was helpful here to remove bits of the branch which were intersecting the grapes. In this case helpful for flexibility more than anything, but the concept is powerful because it doesn't work by changing the actual geometry, its working at a 'raytrace' level, so therefore you can NB anything,

The grape technique is using a thin-sss material, with a lambert refractive backface material, both with additive layers for spec/diffuse.. please note that you should leave the lambert refractive bsdf on its own layer, without any weight-mapping, and avoid detailed mapping of the transparency, by using very blurry maps. This is because you cant texture-map a volume, you can only map to the surface, so keeping the maps very subtle and blurry is the best compromise.

There are some artifacts due to the uv-mapping with this evermotion model (but it is a nice model and the 4 grapes per uv-map setup does make it easy to randomize the look). I could fix it but would need to orient some texture coords for each grape, for the spherical mapping.




Also needed to seperate out the self-intersecting grapes. The easiest automated method was using dynamic tags in zero gravity, with some damping, to nudge the grapes apart, (I had already associated the branch mesh to the grapes meshes, so that the branch polys would follow the grapes.)

Enjoying testing 3.2RC, really like the way its evolving. Have found many stupid little bugs remain still though, esp with older files, and more advanced setups, trying to write them all down.... I hope NL dont release too early..

EDIT: just noticed my plugin was not up to date, so maybe that was causing issues.

[ababak]

Eric, it looks very nice! May I ask you to show the materials you explain? What is the idea behind using thinSSS with backface materials? Basically I don't understand how thinSSS should work on solid dielectrics especially when you add a refractive backface material

Thanks

[eric nixon]

The outer thin-sss material should look like a dry empty grapeskin, I used nd1.3. The drapery preview is good enough to eyeball this. I dont really know the correct values for assymetry and scattering coef, I used -.4 and 1200 for those values. The base layer has a thin-sss bsdf with low roughness which provides the tight specular highlights. I put the diffuse bsdf and broad specular bsdf in an additive layer.

The backface material is pretty much fixed as lambert glass with no diffuse, you can adjust the nd to change the look, I think I used 1.45. I mapped the transparency with a very blurry version of the original grape map from evermotion, and tiled it in reverse to create some variation. After that I added an additive diffuse layer with r2 to add some mapped detail to the edges of the inner grape, but this will barely be visible because its all seen through the skin.

I did one test using the same nd for transparent components on both materials, but I think that caused a slowdown, so its probably best not to do that.

AFAIK its not possible (yet) to use sss for the inside material.

[ababak]

Thank you for taking your time for the explanation!

I've never thought of using backface materials for other than back of the label or the leaf. Now I am playing with your setup but unfortunately I can't see the effect of the backface material. I even use a strong grid in the transparency slot just for testing but it's not coming through that thin SSS shell no matter how thin I make it or how high I make the attenuation distance. I'll have to play a bit more.

[tom]

Very clever and complicated setup, Eric. Maybe the surface speculars could be more complicated and less glossy but it's OK. Have you checked making a small turntable animation with one of these grapes? Sounds like, it may fail in refractions but, I'm not very sure. Just roughly suspected..

[eric nixon]

Hi, I decided to render some (low-sl) simballs of the two materials. The grapeskin looked good on its own but I discovered the lambert glass additive material was glowing, so I dropped the additive layer. I re-rendered the grapes also, now they render a bit quicker and dont glow..., only two hrs this time to get to sl23. (on my ancient workstation 12 core @3.6)

Its nice that we can make things glow if we want to, without much of performance hit. I would half expect a client to choose the glowing one.









Tom, I haven't tried using the correct nd of 1.33, If I do it will look too transparent even with 100% roughness, because a real grape has internal structure which roughs up the light some more. For extra realism maybe nd 1.33 plus some internal refractive geometry would work, but the render time will be longer. Perhaps a thins-sss membrane geo with only slight visibility would do it?

Ababak, I did mention before that weightmapping the bsdf/mapping the transparency only works with very blurry maps. You cant map a surface grid mapped into a volume - conceptually thats not going to work.

The nice thing about using nested booleans for the branch is that I can animate this rig, without worrying about giving the branch any dynamics, the branch will just follow the dynamic bunch-of-grapes, and it will never intersect due to the boolean.

[tom]

Both the core and membrane looks fairly realistic to my eyes. I only think the surface is too much polished And also, how would they look without their membranes? Peeled grapes...

[ababak]

I am just not sure this backface material works that way, that's why I tried to make it obvious for testing. I can't see any difference when I change various backface material properties.

[eric nixon]

Tom, no, I haven't tried the membrane yet, that was just an idea and I am lazy to try it for this model. That render is just the 'correct' non-glow version.

The other problem, to improve the speculars, is the uvmapping - I mentioned this before, I need to rotate the axis of each grape to setup spherical maps, and I cant be bothered

Ababak, it does work, but remember the outside material has much more influence.

[ababak]

it does work, but remember the outside material has much more influence.

Oh, it works for me at last. I needed a lower asymmetry value around –0.9 to make it appear. I'll continue to experiment with that. Thanks.

[ababak]

That backface material shows strange refractions. It doesn't appear where there are no other objects behind.

[eric nixon]

Maybe try it without roughness 0?

Tom, I tried using nd 1.33 and it looks much better, thanks, it didn't become too transparent as I feared,.... sometimes old habits.. etc.

Will post render when its cooked.

[ababak]

Maybe try it without roughness 0?

I tried various thin-SSS roughness settings. The refraction problem still there. The green core appears only where there are other objects behind the sphere:


Maybe you use higher asymmetry setting and the problem is not that visible in your renders? I still can hardly imagine how Maxwell calculates dielectric backface material in this case.

[tom]

Actually, I don't think assigning a rough glass as the backface material could produce a correct look.