Hi, i m trying to render poly object like volumetric cloud using VRayEnvironmentFog. Below is first test output, Currently it looks like 2D texture applied on surface not like volumetric as shown in documents.
So how to do this ?? And share your tricks to render object like volume at rendertime.
The trick like always is the lighting :)… The other important thing for volumetrics is of course the density.
What are the dimensions of the shape?
dimensions = 5, 10, 5
&
Density map is tricky
I could be wrong, but…
Check your remarcher settings.
With a step of 1, you are only getting (at best) 5 texture samples through that geo.
I would set your step size to 0.1-0.5 and see if that helps. You could also reduce your max steps to (furthest point of fog from camera distance) / (step size).
Let me know how you get on.
The step is not much of an issue when you have adaptive sampler since you will get noise which will be resolved with more samples with different random seeds.
Do you mean that no raymarching occurs at all with the adaptive sampler?
I’m basing my suggestions on the fading memory of a project a few years ago (so it’s probably out of date now anyway), but I found taking this approach meant I could get a good result with fog without pushing my image/dmc sample rates to the moon.
No, you always need some form of ray marching with textures. I mean, putting even 4 for the sampler subdvisitions will produce 16 extra rays, which will clear even big steps. Sometimes, it really makes sense to unload the image sampler, but sometimes you can use it’s adaptiveness, since our ray marching is not fully adaptive currently, i.e. it will make all the steps.
If I remember correctly, you can use 1 subdiv in the env fog, and it will fully shade only one point probabilistically (but will still do ray marching to find it). This will make the fog fully adaptive. But of course it has its own overhead…
That’s rather interesting!
I wonder would it be possible to get chaos to do some kind of normal perturbing / turbulence uprezzing stuff so we could use geometry with a procedural displace for the big shapes, then some kind of render time distort / perturb for the finer, silhouette breaking detail?
Would be great to see this lit in more of a skylight / sunlight manner!
The VRayVolumeGrid has displacement that can inject small details. This sounds like a good idea at first, but it’s quite computationally expensive.
Here is a quick test i did, with just displaced geometry and VrayVolumeScatter material.
Still trying to figure a way to get the finer more broken detail. But if you don’t need to go close up I think it works well.
Much quicker render than fog, this has 8 bounces, although you can’t go through the cloud which you could with fog.
I presume it’s because the small details would force the raymarcher to operate at a far finer step value to pick up the detail? It’s almost as if you’d want something similar to Bifrost’s new adaptive grids where the inner, totally enclosed values can be estimated at a lower value and then the edge voxels are calculated accurately or you’d want to have two cloud objects, an inner, simple core cloud and then the outer one with all of the detail. You guys are a very smart bunch of people, I’ve no doubt you’ll eventually come up with something efficient that looks good 
Render time is pain in a**…After increased sub-D from 8 to 32…Got 15 minutes difference…Hope in near future Vray Team will make something like pixar Denoiser.
No, we will not spend any time on that ourselves - there’s enough work writing the renderer itself. However we might partner up with someone else on a denoising solution; there are a few different ways to go, including simply using the Pixar denoiser as a standalone tool.
Best regards,
Vlado