Then I switched to Mudbox for the texturing because I'd never used it for texturing before and I love the fact that texturing works by projecting and not polypainting, like in ZBrush. When texturing, I thought it wouldn't be a good idea to paint microdetails but to maintain the cartoony style of the artwork instead. Because of that I felt comfortable painting and I focused only on giving color variations to the textures. Fortunately I was able to finish painting more the 120 objects more or less in one day.
When I finish painting, I thought it would be a good idea to bake cavity and occlusion maps to use later in the shading process. I used xNormal for the cavity maps and Softimage's RenderMap for the occlusion maps (Fig.08).
Shading & Lighting
When I thought about shading and lighting I knew that I had to make a critical decision... and I made a mistake, which fortunately I was able to fix in compositing. If you look at the original artwork, you'll see that there aren't any shadows. But that doesn't mean it hasn't got volume; in fact it has and it seems like the artist has applied some type of painted occlusion.
Anyway, I wanted to try and achieve that look, so I decide to light the model without using lights. It sounds a little bit crazy, but if you think it through a little, it really isn't. The key is to use a constant material for the base color with some layers (occlusion, cavity, specular) to give volume (Fig.09).
To get specular highlights similar to those in the original artwork, I put a lot of different exclusive lights (without diffuse) around Sandoje (Fig.10).
The mistake I mentioned before is that I didn't realize what the main idea is when you make a 3D image based on a 2D one: you can base your work on the original artwork, but you shouldn't copy it. And that's why later, when I started compositing image, I decided to add an overall occlusion effect to give Sandoje more volume, because without shadows the image looked a bit poor.
When it comes to splitting scene objects into render passes I tend to split them enough in order to have a lot of control when I composite. Anyway in this case I tried not to split so much because I wanted to work with the objects's label channels. This way I could change object colors in Fusion only with one rendered image, but when I prepared and tried this channel in Fusion I found it was really frustrating. I could not find any image format which worked well between Softimage and Fusion.
Because I didn't want to waste more time with it I continued to make it "the easy way. I split the scene into lost of render passes (Fig.11).
Thanks to the baked ambient occlusion maps, I was able to "fix" the lighting problem I told you before, giving the final image more volume. I made custom passes using these maps to work with them later (Fig.12).