I started the Space Bunnies project in August of ’05, planning on a three-month project. I worked out a storyboard and a rough animatic in a week in Flash, and followed up with detailed character and set designs. With those done, I had a pretty good idea of what certain aspects of the the final product would look like. However — I severely misunderestimated three dimensions which pentupled, or possibly septupled, the amount of actual labor involved. Not to blow the surprise or anything, but those three dimensions were BAD RIGS, TECHNICAL SNAGS, and COMP TIME.
I knew a few things for sure: the piece would be five to six minutes long, broken down into 30 to 40 shots; it would have seven separate character rigs and models, five sets, and a number of other rigs for props and pieces of the set. To keep things simple I aimed for render times of 30 seconds per frame, allotting a week for rendering.
At this point some of you may already see my problem: I’m crazy. Maybe so, but not necessarily for the reasons you suspect. True, with a standard single-person setup and workflow, a project of this size would not normally be feasible in three months at anything above stick-figure kung-fu quality, but I had a couple secret weapons up my sleeves.
In one sleeve was a small render farm, assembled from machines I found in the street, with a total of 15 ghz of cpu available. In the other sleeve was a sketch-based free 3D modeler called Teddy, with which I modeled practically everything in the project in a week flat. After painting everything in Deep Paint, I rigged my models and sets, and then set up all the shots with the appropriate sets, characters, and props. I was still on schedule. I was happy! Then I started animating.
Cue Ominous Music
I made a blocking pass, moving things around in rough poses — then I went back and started filling in the details. That’s when I discovered that my rigs were… not good. They were bad. They were BAD RIGS. They fought me like they had rabid mandrills in them. They slowed me waaaay down. I hit my three-month deadline, and I needed my brain for other things, so I took it off the project.
When I came back to it a few months later, I was much refreshed. I attacked the animation with renewed vigor. The characters still didn’t like me and weren’t impressed by my vigor, but we learned how to work together. I began to complete sections of the animation, and to vary my diet and keep myself from burning out I set up my comp structure and started rendering and assembling a few shots. But though my project looked like a piece of cake to the untrained eye, once in the comp it was revealed to be complicated, large, and unwieldy, rather akin to an airliner. And though theoretically one man may fly an airliner, it’s probably best wrangled by a team of specialists working in unison.
Lacking a team of airliner wrangling specialists, I found myself running into TECHNICAL SNAGS — inexplicable and mysterious problems with Maya, After Effects, and my render farm. Most were due to my ignorance, a few were bugs. Some were real monkey wrenches, and the project bogged down again. I pushed myself too hard, trying to power through the pain, and got nowhere. I moved to New York City to relax.
In New York, I learned that all the money in the world can’t help you if you don’t have heart, and that I should fight for my dreams no matter what anyone says, and also I saw Alan Cumming and Cyndi Lauper going to the Tonys and Jude Law in the Village. Jude Law is taller than I thought he would be.
The second time I returned to the project, I was more stoic about my prospects. I’d locked horns with the project already and knew what kind of problems I had to solve. I started methodically chewing on the animation as though it were a really long piece of beef jerky, and bit by bit it became digestible. I kept my lighting relatively simple, and there weren’t many surprises there.
Then I had to render all the shots. To keep things flexible I decided to render in layers and composite them together, so that if I needed to change something I wouldn’t have to re-render the whole scene — I could just render the part that changed. Things looked normal until I started the comp in earnest.
The comps were much larger than anything I’d done before. Even apart from the snags, the amount of COMP TIME involved was a bit overwhelming. Here are some figures I didn’t anticipate:
Average number of render layers per shot in Maya: 5.6
Total number of rendered layers output from Maya: 209
Average number of comp layers per shot in After Effects: 10
Total layers needed in After Effects: 340
Total number of images rendered by the farm: 80,428
Calendar time elapsed from start to finish of rendering: 6 months
Calendar time elapsed from start to finish of comp: 9 months
Now, this wasn’t full-time. I took other projects here and there, and occasionally went to the bathroom. And on the grand scale, nine months is peanuts, and yes, Don Hertzfeldt spent twenty years on his two-minute hand-drawn animation about the universe. But the point is, I didn’t expect it to take so long.
Much of the extra comp time was an artifact of my process, which was tuned to avoid burnout. If I had the fortitude to move a mountain one stone at a time, I suppose I could have finished much faster… as it is, after a few days working on the same problem with no visible progress I begin to wonder whether it’s worth fixing. I begin to think about other things, and run errands. If I smoked, I’d take smoke breaks every half hour.
Faced with 37 shots full of lighting, rendering, and comp problems, I decided to take the piece in passes. For each pass, I went through the entire piece one shot at a time, fixing the most egregious and obvious problems, not worrying about whether the result was ideal. In this way I was able to see real progress, fix similar problems in multiple shots at once, and avoid sinking into a perpetual perfectionist quagmire the way I tend to.
This project took 15 of those passes. I kept my render farm real busy. Were I to render the whole thing all at once, with no layers, it might take a week as I’d estimated. But with layers, and complications, and revisions, it took much, much longer. When the returns were sufficiently diminished, I called it quits, and added the music and the sound effects.
Behind the Music
When I finished Space Bunnies, I started this post. I knew it would have at least five sections, with probably a paragraph or two apiece. To achieve my standards of voice and flow, I took 429 editing passes, and finished a week later. Then I let it “cool” for another week, during which I watched the Pussycat Dolls finale. Then I wrote this paragraph.
I’ll elaborate on the specifics of each stage at a later date, and maybe even draw some conclusions. Right now I want a sandwich, so first I need to make a spreadsheet analyzing the contents of my fridge, which I’ll do as soon as I figure out why it’s throwing rocks at me.« previously: Tekkon Kinkreet | Home | next: Aachi & Ssipak »