Robot Disaster: lessons in 3D animation
Back in January of 2001, I made this animation of a robot.
This was my first experience with 3D animation, and it probably shows. I used Infini-D, a package that was obsolete even then, but easy to pick up. The effort-to-reward ratio was high enough that I was able to delude myself into thinking I had the patience for 3D animation. This turned out not to be true, but the experience was useful: when I saw the trailer for the Jimmy Neutron movie, I could tell based on a certain wall texture that the movie was totally made with LightWave.
I think this animation is pretty decent given my lack of experience, but looking at it closely (using a higher-resolution version than YouTube allows) I’m realizing how much I approached computer animation as an engineer. I’ve never had a good sense of aesthetics, and I thought of 3D as a way to sidestep that: just drop some objects and lights in your scene, point a virtual camera at them, and voila! Instant Toy Story. Needless to say, there’s more to it than that. Even though you’re working in 3D, the finished product is still going to be a 2D image, and you need to have that final image in mind the whole time you’re laying out your scene.
Some of the ways I flagrantly ignored this:
- Nearly everything in the animation is reflective. Mainly, the R on the robot’s chest, the robot’s eyes, and the floor/ceiling. This is dumb for two reasons: one, it’s visually distracting and unnecessary (you don’t need to see the checkered floor in the robot’s eyes, and in fact you can’t — see below). Two, it makes the scene take way longer to render, since everything visible in the reflection basically has to be rendered twice. But I did it because it was easy (just click the “mirrored” checkbox!) and what says “hey look, I’m making pictures with the computer box” more than a reflective checkered floor?
- Excessive camera motion. Again, because I could. Again, distracting and unnecessary. I later learned that moving camera shots, in real life or virtual, should start on something and end on something. Kind of obvious when you think about it, but think about how many features you abused the first time you used a camcorder.
- Details you can’t see. The robot has two transparent spheres on top of his antennae, which fall off when he goes through the floor, but they’re nearly invisible. The green columns have a weird noise texture that looks like TV static when the camera moves. The point where the robot falls through the floor is marked by a particle fountain, but the red particles are hardly visible over the red floor.
- Pointless textures. The robot’s butterfly wings have the same texture as the walls in the second scene, but they don’t look particularly natural on either. The checkerboard floor is kind of a 3D cliche, and when you’re seeing it from a low angle the distant checkers look weird. I’m really not sure why the table has that hideous tie-dye pattern, but I actually did another robot-on-table animation later and the table in that one was even uglier.
- The last scene. I’m not sure if this was a rookie mistake or just a tough thing to storyboard, but switching to a completely different camera angle for the punchline of the animation wasn’t the right choice. I think the surprise still works, but it could’ve been much better.
That said, I did stumble onto one time-saver: robots are excellent for animation. If they move stiffly, no problem! They’re robots! You can also blow them up without being too macabre, and that made Robot Disaster kind of a 3D equivalent of those flipbooks you drew as a kid where a stick figure dude gets eaten by a dinosaur. I’m also kind of pleased with the way that poor robot breaks apart at the end. He’s kind of dumb, too — he sees the end of the table, and he has wings. His destruction is his own damn fault.