[eric nixon]

heads up: additive fix doesn't apply to transparent things (inc thin-sss) but if its dark and therefore the total energy is under 100% it should be fine.

[rusteberg]

Heads up, why don't you post some useful information (visuals work best here) showing how you've intentionally broken shit so we know what you're talking about?

[AlexP]

rusteberg, if you are talking to me, sure I will, I just grabbed topic, as recently I was working on it.

[tom]

Great balloon tests!

[eric nixon]

Sorry guys, I've given-up again on balloons, I feel there are some issues/bugs still, such as strange renders using maxwell-sun, and also slow renders with balloons arranged as a bunch even with soft lighting. Its not so much a material problem, its just the general unbiased render problem with caustics of caustics of ... This layering issue also occurs with leaves on a tree or a grass lawn, or layered curtains.. but with broader assymetry its easier to render (as the transparency is more evenly spread).

I got really excited thinking 3.1 thin-sss was faster in layered situations and had solved the colour inversion 'feature' we see with diffuse+trans in one bsdf), but I was mistaken.

Anyway, heres the last images I saved;





And failing; 2nd image also has bad settings to highlight the sunlight problem.



If you need a white balloon it can be made a bit brighter by mixing diffuse and trans in one bsdf as there are no colours to become inverted, but it seems a bit like a hack, not sure if realistic shading...



So at the moment the best I can do is isolated coloured balloons in softlight (which can be brightened in post via a mask) or white balloons in a bunch in softlight... OR a bunch of darker coloured balloons in softlight.

[eric nixon]

I thought it would be interesting to compare a balloon with modelled thickness VS thin-sss on a thin model, maybe its an unfair comparison but the difference in appearance was surprising, all thats changed between these two renders is to turn on thin-sss (and ofcourse turn off the geometries thickness)



Maybe its a bug, or maybe it needs improvement, or maybe the comparison is really pointless, but I would love it if it really worked... I know thats asking a lot, hire more geniuses?

[eric nixon]

By simplifying the above material, so that it is only 1 bsdf, it renders fairly well on 0.1mm 'thick' geometry, this took 50mins sl20 (3.6ghz x 12c)

And rubber is a homogenous substance which suits regular SSS. With some more tweaking this seems the better way for a still shot at least...