IA: Can you tell us a about the exchange between the different teams while working on a specific sequence? Texturing/ shading / modeling for example.

Neil: The sets team handles everything that has to do with the environment. In the sets team you'll find shading people, modelers and set dressers. How the assets moved around depended on the sequence, the individual artists, and the schedule. For example, I can model and shade, so I got a chance to sometimes both model and shade the same environment. Other times we worked in small teams and one guy would model the lowres terrain and develop the shader, someone else will apply the shader and customize it to the terrain, another person will build more detail into the terrain, then a painter may add some painted detail while the set dresser starts placing the hero pieces of trash in the scene. We also work closely with layout to make sure we know the camera angles for the shots and with the designers who do concept sketches and paintings of what the environments should look like. It's all a big team effort.

IA: What was the most difficult part you worked on?

Neil: Definitely the giant trash piles at the beginning of the film, they needed to look good from both far away and close up for when the camera takes you flying right past them. And it all had to be renderable without running out of memory or taking 20 days for each frame. We ended up using a combination of displaced shader trash, particle trash and hand placed hero trash. Some of the furthest trash was later converted into brickmaps for easier rendering.

IA: Can you tell us a little about the technical aspects and about the modeling complexity? For instance, the number of polys or the computing time?

Neil: It was very complex, which is why we tried to use stuff like instanced particles, brickmaps and displacement shaders when we could, instead of fully modeled and textured pieces of individual trash. But there was plenty of that too; it was placed by our team of setdressers. Also, we spent a lot of time on optimizing the models since there was going to be so much geometry on the screen at any given time. It didn't matter if your model rendered by itself, what mattered was would it render when combined with the rest of the set. Like in WALL•E's truck, there was a LOT of trash on his shelves. A lot of time was spent optimizing every object and reducing the object count wherever it could be reduced. Also, we had a few programmers finding ways to trim the memory footprint of these scenes from a code perspective.