vrayenv fog -struggling to get a white cloud

[super gnu]

im trying to set up some clouds with env fog... i find they are coming out grey whatever i do, even when quite strongly illuminated.. ive got the fog colour set to pure white... however its the transparency thats causing headaches. if i set it to nearer white, the fog gets whiter, but more transparent. pure white, it obviously dissapears. i cant add emission to it as i dont want it to glow.

im guessing some combination of higher density and lighter transparency might get close but IPR is taking forever to refesh as my gizmo mesh is super dense. any tips most appreciated.

[^Lele^]

Have low transparency (start by setting it to black), scatter GI active, scatter bounces to 100 (the same as the LC: not a coincidence, use it. Doing this with BF could cost a wee bit more.).
The bounces are the key, here: the more light is allowed to bounce around inside the volume, the more chanche it has to reach the viewer on the other side.

[^Lele^]

Example on a spherical gizmo set up like so:


1 bounce:

10 bounces:


and 100 (no big difference here due to the gizmo size: light reaches the other end with 10, already.).


Rendertimes were higher (!) for the 1 bounce one, nigh identical, and 30% lower, for the many bounces ones.


To render convincing clouds we'd need phase, though, so perhaps you should try a different approach, with one of the volume techs which support it? (f.e. volgrid, scatterVolume)

[super gnu]

im trying with scattervolume, but im finding, randomly, that an object with scattervolume on a) doesnt show up in multimattes based on matID and b) ignores light exclusions.... both of which i need for my scene.

[super gnu]

also, out of interest, is there a way to "cut" a scattervolume with an opacity map like there is with sss2? some kind of blendmtl setup? ive tried everyting i can think of but didnt find a solution.

[super gnu]

for reference, here is what im attempting:

earth clouds... here is a "crap" version with a simple material on opacity mapped sphere:




here is a version using sss2, and opacity mapped to get detail... for the volumetric ones i generated an insanely dense displaced sphere using max displace, added an inner skin so it was closed, then proxied it out.



here is a version using scattervolume:




since it cannot use an opacity map, it looks more "blobby" since its just the resolution of the displaced mesh defining the detail.


the sss2 version works the best, by far, but it takes an absolute age to render, and i have 4000 frames to do. . possibly made worse by the opacity mapping i expect.


the scattervolume is much quicker, but looks less detailed and has the aforementioned problems (no multimattes and no light exclude, both essential) .


ive also tried, as mentioned above, with the same mesh as a gizmo for env fog... that also, kinda works, and is still faster than sss2, but im not using GI, and its always grey, or transparent. ive tried enabling GI but it has very little effect and kills rendertimes too. if i could create a multimatte for it i could fix that in post... but the whole atmosphere is inside another env fog, so i dont see how to generate a mask.

[^Lele^]

At this distance, perhaps you can get away with clip opacity?

[super gnu]

there is no "clip opacity" option in sss2.... also i do have to get a bit closer.... and the map has many shades between solid cloud and no cloud....

[Joelaff]

Could the render element issues for the mat ID be caused by another outer refractive object not set to have refraction affect "All Channels" ??

Did you try using a multimatte element and selecting the scatter volume object in the include/exclude vs.using isMatID ? I think that still needs any refractive objects between it and the camera set to All Channels.

Could you use VRayDisplacementMod to get the detail in the displacement (higher render time poly count) without it being blobby?

[Joelaff]

Could also try old school fakery with multiple clip mapped spheres of slightly different scales. Not going to give you proper scattering per se, but would be faster, and might work depending on how close you get.

[^Lele^]

there is no "clip opacity" option in sss2.... also i do have to get a bit closer.... and the map has many shades between solid cloud and no cloud....

Agh, ofc.
Then yes, it's likely to be normal opacity, so costly.
Have you tried the standard vray material at all?

[Joelaff]

Have low transparency (start by setting it to black), scatter GI active, scatter bounces to 100 (the same as the LC: not a coincidence, use it. Doing this with BF could cost a wee bit more.).
The bounces are the key, here: the more light is allowed to bounce around inside the volume, the more chanche it has to reach the viewer on the other side.

Wow. This ought to be in the docs more clearly like you present it here.

Though the docs have some samples they do not make clear the utility of multiple scatter bounces. Also the default of 1 looks pretty terrible for everything in comparison. I wonder if the default should be the same as the light cache.

[^Lele^]

It's very easy to run out of luck combining bounces and textures, and hence raymarching.
That's why the defaults are as they are.

As to the specific setup, it makes logical sense, if you think about it: if the volume is transparent, then it lets light leave before it had a chance to lit it.
If you make it more opaque, once light is in, it can't get out as easily, and it ends up lighting the volume instead.

[super gnu]

Agh, ofc.
Then yes, it's likely to be normal opacity, so costly.
Have you tried the standard vray material at all?

ive never had luck controlling the translucency settings in the standard vray material.... i should probably revisit it.

regarding opacity, on this job ive actually avoided it like the plague, using refraction mapping with ior1 wherever id use opacity.. its much faster than opacity, for some reason, given it produces a visually identical result, something ive never understood!

[Joelaff]

It's very easy to run out of luck combining bounces and textures, and hence raymarching.
That's why the defaults are as they are.

As to the specific setup, it makes logical sense, if you think about it: if the volume is transparent, then it lets light leave before it had a chance to lit it.
If you make it more opaque, once light is in, it can't get out as easily, and it ends up lighting the volume instead.

The theory makes perfect sense.

What do you mean by "run out of luck" above? The defaults generally produce useless results for me. I am typically after thin, illuminated fog or rays, and almost always have to use an empty VRayVolumeGrid, and I think the default of 1 scatter is the likely culprit.