Did some experiments with volumetric shaders today. With exception to light every single parameter was generated with a map that “the shadowland machine” created.

Version 1

high volumetric density. colors and displacement from a shadow map

Version 2

lower density, colors and displacement from a shadow map

Version 3

density also driven by a shadow map (as usual: colors and displacement from a shadow map)

Quick summary: I am not happy with the outputs so far. The images look cheesy and kitschy in a bad way.


Version 4

Idea: Multiple iterations of displacements. (“Staging and layering-approach”). Use a shadow map as base mesh displacement and a second shadow map as volumetric displacement.

looks a lot better. Unfortunately the render times rise enormously: Full HD image with 50 samples: approx. 3minĀ  on GTX980ti.

Version 5

Closeup version of the cloud. Maybe render a small camera flight through the volumetrics?

I like the abstract qualities that are possible with this image.

Update (13-03-2022)

Since the rendering did consume too much time, I activated a bigger machine (switch from GTX980ti to RTX3080) and rendered out a small animation. The switch cut the calculation times by approx. 70%, although it is far beyond any real time capabilities.

I animated two domains:

  • the amount of volumetric displacement (from 0 to 0.5) = i.e. the influence of the 2nd map on the volume (I did not animate the map itself)
  • a small camera movement