John Gruber linked to a great article the other day about the steps Pixar took to improve the realism of Wall-E. The problem with computer-generated movies is that the computer perspective looks so perfect, the creators need to put some mistakes in to make it look more like a normal film.
One aspect I found interesting was separating the tilt and pan location from the position of the lens. Obviously these are separated by some distance in a real camera (the camera’s centre of rotation is somewhere behind the lens), but previous Pixar hadn’t done this in their films:
On the old Pixar camera, the tilt and pan point of rotation was right where the lens was. So any time you did a tilt or a pan, it was just a 2D move. And Lasky says it looked like you could do it in After Effects — there wasn’t any change in perspective.
“The first thing we did was move that away so it felt like where the camera was rotating from was different from the lens. It’s one of those things that’s very subliminal, but any time you moved the camera — and we moved it a lot — it felt like it was actually in the space rather than looking at a screen through something. That was a big change. So many camera packages out there work that way, but we had never taken that step.
I’m quite keen to see this movie. It looks like another great one from the guys at Pixar.