The first day of SIGGRAPH 2008 has passed and it has been quite overwhelming. There is so much going on at every hour of every day. Classes, talks, tech demos, festivals, and everything graphics related. I’m already exhausted from today, but now I know what to expect for tomorrow and the days to come.

The day started off with Joel at registration and then off to the merchandise pickup for some swag. Our first item was to attend a class on Advances in Real-time Rendering in 3D graphics and games. Most of the material was way over my head but I did catch some vocabulary words I had recognized. It seems like spherical harmonics were a pretty big thing for real-time lighting as that came up numerous times.

AMD was probably the most interesting talk of the morning. They talked about ways to simulate crowds using only the GPU while still maintaining reasonable visual quality. Their demos and technology were fun and engaging. Sadly, I couldn’t say the same for all of the talks but that’s probably my lack of knowledge, not their lack of clarity. This class was just the start of a long and busy day…

Ed Catmull: “Managing the Creative Environment”

I really wanted to see Ed Catmull’s talk and so did most of the attendees at SIGGRAPH. The line (which was more of a clump) was already growing 30 minutes before the talk. Apparently there is no back door to the stage, because someone tapped on my shoulder and said, “Excuse me.” It was Ed Catmull. Oh my god, he tapped me! I turned to see that I had been standing next to Rob Cook the whole time as well! While I’m at it, I also saw my old cs248 professor, Kurt Akeley walking around in a cool cowboy hat.

“Story is King”

Catmull started by talking about misconceptions about what makes a studio a successful creative environment. He challenged the belief that “story is king” as the recipe for success. After all, everyone believes this, and there are plenty of examples where this vision is simply not enough. He says that statements like these that seem to have some insightful meaning are really just tautologies and provide no substance.

He asks us to keep two questions in mind:

1. Is the problem the absence of good people or good ideas?
2. What does it mean to have a good creative community?

I think he only addresses the first directly, and the second is kind of just skirted about by digging in deep to the inner workings of Pixar, which is, by definition, a good creative community. He resumes his talk, focusing mainly on the body of talent that seemed to gather around Pixar at the start. He talked about the initial group at Pixar that was so driven to making a full-length computer animated film. He traced Pixar’s history as it moved from New York Institute of Technology to Industrial Light and Magic. Pat Hanrahan’s name came up as he talked about Pixar’s software (relating to Pixar’s manufacturing days). It seems like the point of all this name dropping was to conclude that it is the absense of good people, not good ideas. The prescence of the big names associated with Pixar and Pixar’s remarkable success is no coincidence.

Standards of Quality

And then, there was Toy Story. Admist the success of this film, another group began working on the sequel, Toy Story 2. Before then, the only computer animated sequel was Rescuers Down Under, which he claimed was not very successful. I couldn’t disagree more as it is one of my favorite animated films! Anyhow, Catmull said that early on, the quality expected from Toy Story 2 was not the same as with Toy Story or A Bug’s Life. They realized that this was a huge mistake and John Lasseter had to step in the last minute to correct it. The lesson they learned here was not to have different standards of quality for people working together.

He also inserted here that the main story behind the original Toy Story 2 and the one released were both the same. But having the expertise of John Lasseter and others that worked on Toy Story made it into a masterpiece. So, it goes to say that mediocre people with a mediocre idea will fail. However, great people with a mediocre idea will either fix it, or drop it.

Pixar’s Workflow

The rest of Catmull’s talk was about Pixar’s pipeline and what sort of quality control they install to produce hit after hit. The principles are not too different from those of software development. He stresses the importance of iterative development:

The first time, it sucks. If it didn’t, it’d almost be done.

Pixar has daily’s where everyone showcases their work and coworkers are free to openly critique one another. He says this helps artists break out of their comfort zone (who often are embarassed to show incomplete work), enabling them to be more creative.

Pixar also have post-mortems. From the film, The Pixar Story, they also talk about the fear of complacency after being so successful. That was the reason they brought in and outside director Brad Bird for The Incredibles. Catmull says:

Things that go right are excuses to not look at what is wrong.

They found that the most productive way of performing these post-mortems was to have each person say the 5 things they’d do again and the 5 things they wouldn’t do again. It is a fair sampling of the good and bad lessons from a movie.

The Brain Trust

Pixar has what is known as “The Brain Trust.” I’m not entirely sure what the purpose of this was but it seemed to have most of the creative giants of Pixar. The Brain Trust had three rules:

1. Only filmmakers. Ed Catmull was not part of The Brain Trust.
2. Must be honest, completely honest. Most people were afraid to be honest or didn’t want to be rude.
3. No authority over directors. This rule added a lot to the dynamic as directors knew that suggestions from The Brain Trust were in the best of interest of the film and nothing else.

Overall, it was a pretty engaging talk and he brought up a lot of interesting ideas. I think it was a bit hard for me to follow some of his points, but it’s always great to peek inside the inner workings of great companies.

Starcraft II

In the afternoon, I attended the second session of the Advances in Real-Time graphics class. I fell asleep during the first session where he talked about wavelets to do adaptive compression of textures. Most of it was incomprehensible to me anyways.

Next up was Blizzard where they talked about their up and coming game, Starcraft II. They had two speakers, the lead engineer and the lead artist. I really liked this format. They would tag team with one another. First the artist would show examples of things the artists wanted to accomplish and then switch over to the engineer who described how they went about it (or compromised).

The talk focused around screen-space algorithms and the different effects that could be achieved using 2D + depth instead of full 3D. The main technological points hit were:

1. lighting,
2. depth of Field,
3. and translucency and translucent shadows.

They made a lot of approximations but showed that most of their results were sound. I liked this engineering approach that Blizzard has taken. It seemed like a lot of studios want to solve things physically correct which is an admirable goal; however, Blizzard seems primarily focused on making kick-ass games. Instead of going for realism (which wouldn’t be necessary in a real-time strategy anyways), they go for accessibility (apparently it runs on Radeon 9800!) and gameplay.

Fast-Forward Session

The last event of the day was this Fast-Forward session of the many technical papers to be presented at SIGGRAPH. Each paper has 30 seconds or a minute to pitch to the audience in whatever way they want to convince you to come to their talk. A lot of them were funny and I will be going be going to these rather than classes. One strange bit was the computation/numerical analysis groups’ fixation on killing the Stanford Bunny. Sadness.