Working from this sketch Andy supplied, which communicates camera movement, contrasts in the scenery and anything that might warrant some problem solving. Of course, the first thing that comes to mind is practicality of our solution.
There's only so much time and lots of sets to be done and only some of the stock material will provide nice negative shapes (it's mostly used for texturing).
Going to try:
• Subdividing in maya and adding a texture with a noise displacement map (839,263 faces to 3,357,052)
• If that looks like crap, gonna subdivide in Z-Brush once more and refine
... time passess...
Well, not enough, still. What I really need to be able to do is have a base plain field of what looks approximately like ash that I will build photo-details and animations into.
I've put together something out of regular Billow noise and Worley noise to get a grain.
Pretty good. Like a desert. Worley at frequency 256. Trying 1024 for the full-size render, just hoping to capture a similar grain. Next step, on top of all this, is a noisy bump and displacement map.
A lengthy render later, the uniformity of the noise map works really well. Unfortunately, it introduced new problems- specifically pixelization on the hill in the middle and a weird lighting artifact that creates a vertical section that looks like it just wasn't comped in well. Also not something we want, so the original map was already better.
Trying a different lighting setup to see if it might've been a fluke or to see if the banding problem in the middle is because of the texture. Re-rendering with a light coming northeast instead of southwest simply bands everything right of the line above and presents horrific artifacting in the tallest hill. Although this is first and foremost because of poor topology (the faces on the tall hill are too stretched) so let's try to solve that first.
Subdividing twice to 1,887,608 faces eliminated the pixelation and banding, but without textures. RM wasn't handling it for whatever reason. Time to move off it.
This work breaks down to some 40% waiting on renders. Use of CG is very useful for the background, because hand-painting an ash field at 16896x4096 is impractical (and inefficient), particularly with the desire for multiple plates of this size.
Ash is just a lot of noise of varied size. The word 'moon' came up a lot in discussion of these scenes, so I started approaching it how I'd approach a moon scene. I'd do some quick sculpting and use a displacement map.
Ash is basically just a lot of particles that are really light and start floating at the slightest touch.
The density of the noise (and visually, ash is just noise) is what sells the scale.
The intent is to show multiple characters popping up from under the ash while panning in an epic sized landscape.
The computer has a finite amount of memory, thought plentiful under normal circumstances, they're a bit shorter when trying to generate repeating 4k maps to help sell the illusion.
So I made a collision mesh by reducing the original mesh's verts from 1,898,165 to 7702 and placing it in close proximity, otherwise it'd never run..
My plan is to cover the environment with particles, which will help suggest scale. Of course, nothing goes without a hitch. Renderman kept refusing to use the camera I selected and gave me grief even after I deleted all cameras - it would only render a small portion in the middle.
A while later...
Somewhere along the way my last mesh lost all its UVs. No matter, it was time to make a new mesh anyway. So I did.
I'm going to compose it out of two different shaded images. One for creating complex texture where I need it, one for the general color/grading base.