Daniel Levitin reveals how to cope with information overload; Polly Morland delves into the subject of risk
I’ve always been interested in maths and physics. In graduate school I was drawn to computer graphics because it made those disciplines really come to life in a visual way. Pixar’s President Ed Catmull was on my PhD committee and it looked like Pixar was going to be a really interesting place for the next 10 or 20 years, so that’s how I ended up here.
We find the problems that are too risky for other people in the company to tackle: either because they will take longer to solve than a film can afford, or because the probability of success is so low that a producer isn’t going to want to fund the project. As far as I’m aware, we’re still the only studio with a research group that’s divorced from particular films. We can focus on those potentially game-changing technologies.
One of the biggest gambles is to do with our lighting artists. For years they’ve worked in a fashion that requires them to move a virtual light source and then wait hours for the new image to be produced, because of the computational time of rendering. We have a research programme whose goal is to make that fully real-time. They’d see it like a physical light, as fast as it would move in the real world.
It’s completely different. When I got here they had just finished Toy Story and were starting on A Bug’s Life. They made Toy Story by the skin of their teeth: the humans were not nearly so compelling as the toys. The world and visuals were pretty simple. At that time, we couldn’t put most things directors envisioned on the screen; we just didn’t know how. Now, we’ve learnt – either through technology, artistry or a combination of the two – that whatever a director envisions, we’re very likely to be able to get it on screen, somehow.
Previously, surfaces were created by two methods. One was flat polygons like you might see in computer games. Those are nice because they are very flexible. But for film we want to create the illusion of smoothness. To get that you need a tremendous number of polygons, which becomes unwieldy.
The other method allowed mathematically smooth surfaces to be used, but they could only be pieced together in very simple ways. To model Buzz Lightyear’s face in Toy Story they used one surface patch for most of his face and a separate patch for his nose, which led to a sharp crease between the two. At a few places in the film, you can see the creases.
What subdivision surfaces do, is let you model with polygons but then turn that collection of polygons into a mathematically smooth surface with no seams. The artists don’t have to worry about them – they just go away.
The best time to identify problems is the middle or late story process, and very early in pre-production. We scout the reels and look for things that might be challenging. For The Incredibles, researchers got involved as the story started to solidify. They worked on methods to make the characters look more volumetric, like they had an underlying musculature.
Are the calculations in animation physically accurate? For us, Newtonian physics is a sort of rough guide: we often deviate from real physics to achieve a heightened creative effect. Friction plays a huge role in how actual clothing behaves; we have models of friction that aren’t physically accurate, but that give us a lot more creative control.
Which Pixar films do you get a kick out of watching? Toy Story 2 scores pretty highly for me. It was the first feature film I really had a lot to do with and it’s also just a terrific story.
The medium is almost infinitely flexible. The look we have now is not quite hand-drawn and not quite photo-real, but somewhere in the middle. There’s still the potential for radically different styles that we haven’t explored much.
My kids and I build things in a workshop in the garage. We made an animatronic dragon called Saphira that blows fire. We basically learned everything about how to build a robot that would appear at Disneyland.