Computer Graphics World

June 2009 16 n n n n Character Modeling•Animation ware, shaped the mesh topology created by the modelers to match. "We set up the mesh so the rigging controls created the shapes the animators wanted," Jordan says. "But in the case of Carl, we spent quite a bit of time and a lot of trial and error to give a square the ability to have human emotions that are not symbolically square." One problem was Carl's mouth, which stretches in a hard line across his huge bot- tom jaw. "We had to make some compro- mises," Jordan says. "We softened the cor- ners, but we tried to do it in a subtle way so you would feel it without seeing it." Cranky Old Carl Animators working with Carl discovered that economy in motion matched his sim- ple style. "We'd put his face in a pose and let the pose speak for the character," says Scott Clark, animation supervisor. "Carl feels more alive if you don't animate him than if you do. Sometimes, we'd just put a blink on him and not much else. He's old. He doesn't bend or twist too much." Animating Carl's clothes, however, be- came a technical problem. To give view- ers the feeling that he lives in a miniature world, the designers increased the size of the weave on Carl's herringbone wool jacket to look as if someone had scaled up a doll's jacket to fit someone five feet tall. "And, it isn't just the textures," says Jordan. "It's the way the fabric moves, as if someone cut a doll's jacket from a life-sized jacket. But nothing in real life looks like a scaled-up doll's jacket. So we had to guess. en, we worked with animation to do walk cycles, have him sit and stand and interact with Russell to see if he was believable and fit the design of the film. We didn't want him to feel like a doll." In the end, Carl looks like a little old man in an oversized suit, a square, boxy shape with wide pant legs. And that caused problems with cloth simulation. "We wanted the miniature look with the over- sized clothing, and to see the lines of ac- The effects department's role was to create the believable elements that connect the audience to the stylized world. "By animating the balloons in a realistic way, the audience more easily believes the house could float," maintains Steve May, supervising TD. "When they buy that concept, they believe Carl is going some- place. And, when Russell shows up on his front porch, they believe he really is in peril. Our primary goal is to help tell the story. Our secondary goal is to make something beautiful." At first, knowing it would be difficult to run a rigid-body simulation on 10,000 balloons, Pixar tried to float the balloons procedurally within a modeled canopy. "They had no notion of each other, though," says Gary Bruins, effects supervisor. "And, they intersected all over the place. It didn't look good." TD John Reisch did tests using rigid-body dynamics with Open Dynamics Engine (ODE) through Autodesk's Maya that produced the believable motion they wanted, but the sys- tem was able to simulate only 500 balloons. When Eric Froemling split ODE away from Maya, though, and created an improved, stand-alone version that ran in Pixar's pipeline, they upped the number. Considerably. With ODE installed in Pixar's pipeline, Reisch created a system in Maya that generated the initial state of the balloon simulation, then moved that state to ODE as a stand-alone without using data communication. "Releasing it from communicating with Maya released the bot- tleneck," says Bruins. "We could do 50,000 balloons, but because they weren't intersecting, we needed only 10,000 to fill the canopy." The canopy provided artistic control for the overall shape of the grouping. "The simulation can push the balloons out of that shape, but when the rigid-body sim was in its rest position, it had the same topology as the canopy," Bruins explains. Thus, to maintain the art-directed shape, the team decided to move the strings that tied each balloon to Carl's chimney in a separate simulation. "The strings are more complicated than the balloons because getting them to bend and conform around the balloons requires more control points," Bruins says. "If all the strings had to avoid the balloons along their paths, they would become tangled and twisted." So, in the first pass, the crew simulated the balloons with strings that didn't know about other balloons. Then, they brought that simulation in as baked data and simulated only the strings. "We thought about reducing the number of strings, but as it turned out, we didn't have to," Bruins says. "The strings col- lide and interact with surrounding balloons, but they don't influence the behavior of the balloons." For close-up shots, though, they would simulate balloons and strings together, and hand dress the hero balloons. –Barbara Robertson Beautiful Balloons

• Cover