Started as a study on snowy mountains using the z-component of the normal vector. Found some annoying cross-talk between normal bumps and displacements, avoided by not using the normal input in the layer weight node.
Transitioned the landscape to a castle scene, which then became practice using render layers as the vertex count grew. Some strange and maddening streaks of alpha transparency appeared in compositing that went away when the defocus and glare nodes were wired in parallel instead of series. Also, the z-pass input seemed to be scaled oddly and did not act as expected with the defocus node (will look into that some more).
The snow on the castle looks like cowhide instead of snow. Have you considered a night scene? Maybe the top of the castle could be a beacon of some sort.
Jagged mountain faces would be good but for my piece, I’d say rotate the small spires because the snow patterns are 100% identical. My eye went right to them. I think the snow is piled and wind blown quite well on them. Another suggestion for scale would be some hint of an entrance or rocky stairway. That tree was a smart addition. Foregrounds help backgrounds a lot.
See that. The spires are already rotated wrt each other. Might have to mix up their textures a bit.
There are actually two figures walking up the left slope to the castle. I’ll try to bring them out more. Maybe add more trees to bring the tree line up to a believable level closer to the foreground tree.
