• Blogs
• Tradeshows
• Partners
• White Papers
• Events

Oscar's New Clothes

Mar 1, 2006 12:19 PM, By Ellen Wolff

Recognizing 20 Years of Cloth Simulation Achievements

 Digital Content Producer's and Millimeter's coverage of past and present award nominees/winners

Oscar, the famously nude symbol of the Academy of Motion Picture Arts and Sciences, won't be wearing smart new outfits anytime soon. But one clever CG trend that was clear this year at the Academy's Scientific and Technical Achievement Awards was that clothes — virtually speaking — can make the man.

While CG clothes once resembled rubber suits, or appeared to be painted on a character, today's computer-simulated clothing is convincing enough to work in a wide variety of films. From stylized characters like Mr. Incredible to Spider-Man's digital doubles, good cloth simulation is having an undeniable impact on moviemaking. So it's not surprising that the Academy has honored the technical “tailors” behind this trend.

But what might surprise some is that these honors were almost 20 years in coming. Cloth animation based on physical principles attracted CG brainiacs long before the technology existed to make these ideas practical for film production. So when the Academy's Sci-Tech committee decided that the time had arrived in 2005 to honor the digital cloth techniques developed at Pixar by David Baraff, Michael Kass, and Andrew Witkin, it also reached back in time to honor what Sci-Tech Committee member Ray Feeney calls “the original ‘Aha!’ moment.”

The First Cut

That moment would have been in 1987, 10 years before Pixar wowed CG fans with the clothed human star of the Oscar-winning short, Geri's Game. Back then, computer scientists Demetri Terzopoulos and John Platt were working at the Schlumberger Laboratory for Computer Science in Palo Alto, Calif., but light-years from Hollywood. They attempted ways to simulate moving, deforming objects. They reasoned that they could model the shapes of flexible objects and apply simulated physics to govern the movement of these objects. (Think flags flapping in a simulated breeze.) Platt, now a Microsoft researcher but then just a graduate student, recalls that the prospects were daunting. “We knew it was difficult,” he says. “Even with a computer like the Cray — the fastest at the time. But we just got short simulations.”

Yet this pioneering research — the “Aha!” moment — led the Academy to honor Platt and Terzopoulos with Technical Achievement Awards decades later. “It is surprising after 20 years,” says Terzopoulos, now the Chancellor's Professor of Computer Science at UCLA. “But on the other hand, I have seen how cloth simulation technology has evolved in motion pictures, so I know that our work has had an impact.”

Terzopoulos recalls that, as early as 1988, the idea of virtual clothes was evolving. Fellow researcher Kurt Fleischer (now at Pixar) was able to model a robe that was draped on a mannequin. So, Terzopoulos notes, “We always had the idea that this would be useful for clothing.”

Reality Checks

But making the process useful for production was another story. For example, being able to control all the collisions that occur when cloth moves — like between characters and their clothes — was hardly a trivial challenge. That would require much more advanced software and more powerful hardware. As Feeney observes, “Some of the work from these early pioneers was indeed the right approach, but it was computationally too expensive for the resources at the time. Its practical implementation in moviemaking had to wait until computer resources could catch up.”

Even by 1995, when Pixar made its debut feature, Toy Story, animators were still creating every cloth wrinkle by hand. Director John Lasseter and Pixar's CTO (now President) Ed Catmull knew that if they wanted to animate stories featuring humans, they would have to address the issue of clothes.

Proof Of Concept

As Pixar had done before to solve technical problems, a short film was commissioned. Under the direction of Jan Pinkava, it would test ways to animate clothed humans within the context of a production. Pinkava's 4 1/2-minute film, Geri's Game, starred a man with more convincing skin and clothing than audiences had seen before. It was the first indication that cloth simulation techniques were indeed becoming ready for prime time.

Platt says, “[Achieving this milestone] took farsighted individuals. Pixar devoted enormous resources toward creating the complicated algorithms needed to get Geri's suit to crease the right way when he moved to a new position.”

A key person who made this happen was Kass, a senior scientist at Pixar who is likely the first person on a production team to earn the credit “cloth dynamics.” The simulation that was achieved for Geri's Game opened up a major area of research at Pixar. Ever since 1998, the work of Kass and fellow Pixar scientists Baraff and Witkin continued to demonstrate how simulation could be used efficiently in production. In honoring the trio with Scientific and Engineering Awards this year, the Academy's citation noted, “Their work provided the conceptual foundation for many cloth simulation systems in use today.”

Common Threads

Geri's Game provided a high-profile proof of concept, but it also reflected a larger story because, remarkably, Kass and Witkin had also worked at the Schlumberger lab where Platt and Terzopoulos had done their early research. Witkin recalls, “We weren't working on cloth at that time, but on techniques that could apply to cloth. So there is a common thread.”

Witkin would continue to pursue his interest in cloth simulation at Carnegie Mellon University, working with Baraff. The ultimate outcome of their research would become Alias' Maya Cloth software, used on movies like Stuart Little and Spider-Man. Terzopoulos says, “Cloth simulation started having a broad impact when it got into software packages that artists could use.”

Baraff observes that a rather curious phenomenon had happened between the earliest research into cloth simulation and its popularization via Maya. “When Demetri and John did their work back in the late 1980s they had the right idea. But because computers were not fast enough, their idea kind of got lost. People actually forgot their techniques and started doing simpler things. Once computers got faster, their ideas had a resurgence.”

A Monster Challenge

While digital artists began using Maya Cloth software in a variety of pictures, studios with R&D departments continued to develop unique variants of cloth simulation. Pixar's cloth simulation software, which is proprietary, has become faster and more robust in the years since Geri's Game. Kass says, “We developed a simulator that has more control of the process.”

And they'd need every ounce of control and efficiency they could muster to apply cloth simulation to Monsters, Inc. That 2001 feature would be Pixar's first attempt at doing simulation on a massive scale. Compared to the 89 simulation shots in Geri's Game, Monsters required hundreds. The technique was used not only to simulate costumes but also the flowing hair of the lead character Sully.

Cloth simulation went fairly smoothly during Monsters thanks to algorithms for untangling clothes when, as Kass says, “they were excessively tortured by animators.”

Animators still wanted the latitude to focus on performance, and not on worry about whether simulated sleeves would bunch up. Witkin says, “There was no way we could dictate to the animators what they could or couldn't do. That meant developing techniques that would stand up to pretty much anything they could throw at it.”

Custom Fit

Pixar's simulation designers knew that the studio's animation pipeline would fundamentally change in order to incorporate dynamically simulated clothing. Artists first had to send an animated body to the simulator, which would then “dress” it and make the clothes match the body animation. The software would compute the physical forces acting upon the clothes to make them respond believably. This had to be accomplished with as little human intervention as possible so the pipeline wouldn't grind to a halt. Even though simulation added an extra step to the process, Baraff asserts, “It didn't create a bottleneck.”

“[It did require] a cultural shift to get animators to use simulation at all,” Kass says. “If you want them to give up control of something like clothing, you've got to convince them that it's going to be better. They're willing to give up control of the rendering, because they've become convinced over the years that computers can do a better job than they can do with a paintbrush. But it was a long struggle to have them give up control of clothing. We had to convince them that it would be better than what they could do themselves.”

While simulation provides complex cloth movement that artists could never animate by hand, there were sacrifices. Animators couldn't ignore parts of the character that the camera wouldn't see, because the simulator used unseen body parts to determine how clothes would move. Kass says, “Animators have spent their lives animating to the camera and now we have made them pay attention to whether body parts are going through each other. We thought ‘Maybe if we just reason with them, they'll stop doing that and give us physically possible motion!’ Eventually, what we learned is that animators sometimes need to be able to do extreme poses — even if body parts pass through each other. So we resolved that we had to make the simulator as robust and resilient as we could.”

Incredible Lengths

That was certainly needed for Pixar's 2004 release, The Incredibles, in which director Brad Bird applied the squash-and-stretch animation tradition to 3D, costumed humans. The simulator had to tolerate physically unrealistic behavior and still keep the clothing from getting tangled up. When errors occurred, the simulator had to repair the cloth and restore it to its expected form.

The clothes had to fit the woman, however wild her movements might be. Baraff recalls, “We had very long discussions about what Elastigirl's bathrobe would do when she would reach across the room!”

Feeney says, “[Cloth simulation] came to the fore in The Incredibles. The costumes were character points. Clearly there's an impact of these tools on storytelling.”

Next Year's Models?

What happens next with simulation technology is open to experimentation. Witkin says, “There certainly are more steps that we can take to improve it. And there might be some pieces that we would review to see if there's a better way to do it.”

Baraff agrees, noting that there will likely be new challenges as the resolution of feature films rises. “We're still scratching the surface,” he says. “As our confidence grows, we'll start to simulate more things that are closer to a character's face. The camera spends a lot of time focused on the face, and if you have complicated cloth up around it — like collars and lapels — it will take another level of daring to simulate those.”

“A huge step forward,” adds Baraff, “will be much more controllability than we have now. I remember a director saying, ‘That looks great. But could you make the cape look more dapper?’ I knew exactly what he wanted, but how do I translate that into a set of controls? I don't have a ‘dapper’ knob!”

Cloth simulation still may be a work in progress, but the Motion Picture Academy's Sci-Tech Committee was certain that the ideas of Baraff, Kass, Witkin, Platt, and Terzopoulos have had a proven impact on moviemaking. Feeney says, “There was a great deal of consensus about who were the towering figures in cloth — it's these guys.” And Feeney, like most film fans, expects to see further dazzling developments. “Computers get faster and faster,” he says. “Moore's Law is on our side. So by no means are we done yet!”

© 2008 Penton Media, Inc.