Hey guys, I've had some time this weekend to try out some groundcover.
The initial plan was to model some trees, but that just takes a lot of tinkering in onyx and I just wasn't up for it this week. So I took it out on the grass.
First, I tried to make my own grass objects, you know: take a plane, apply some bend and twist modifiers, scatter, on a small plane, collapse to single mesh and convert to proxy.. then scatter the proxys on my ground plane. Took a lot of time with disappointing results. Soon I realized all my grass blades were all rooting vertically and then bending. Also the small patches of grass didn't quite cover the ground and I had to raise the proxy count substantially or even paint in the missing patches. It was also horrible on rough terrainn, cause the patches were a bit too big. For the type of grass I was going for (also for close up renders), that just didn't cut it. I tried adjusting that, the results were a bit better, but it still didn't look right to me.
Then, I started searching for some free grass models, that had all the properties, that I was looking for. Those were multiple angled roots, non-circular or non-square distributed models and lots and lots of variation. I found some nice free sample models from Maxtree's MTgrass pack.
The result was nice, I converted the Vray materials to Corona and adjusted them to fit the shader I was looking for. I also modeled some clovers in Cinema4D (I'm a max user for about a year now, so I model faster in Cinema:)) and scattered them using a greyscale map.
The distribution and variance looked way nicer, but mostly flat. I then downloaded two more groundcover models from Xfrog samples site, and redid the process, adding those in.
The final step was to add some rocks. I "modeled" those using the greatest debris plugin out there, DebrisMaker2. I used a single texture, spherical mapped in diffuse and adjusted the reflection and bump a bit.
The shot is very closed, because I'm just using a small ground plane for testing purposes.
The main realization with these tests, (as I initially presumed but was trying to avoid and take shortcuts) was that it's equally important to have or make good, varied grass models, lots of them, distribute them properly, and also have good shaders...
A lack of variance in any of those aspects just doesn't deliver the required results.
It was rendered in 5 hours at 50 passes using a single Intel i7 920 at 2.8Ghz @ 3000x1600px with DOF enabled. I just made some basic color correction using levels and curves.
I'll apply the tested grass to my previous Ivy scene in the near future.
As always, any comments and critique are very welcome.