Issue link: https://digital.copcomm.com/i/639267
16 cgw j a n u a r y . f e b r u a r y 2 0 1 6 Believable Destruction Effects TD Supervisor Daniel Pearson led a team of 24 artists in ILM's San Francisco studio, four in Singapore, 14 in Vancouver, and four in London who created fire, water, smoke, dust, explosions, and tons of destruction – ships blown apart, buildings collapsing, planets exploding – for Star Wars: The Force Awakens. "On this project, fire, water, and smoke were grid-based simulations us- ing proprietary soware based on the FLIP solver. We used our proprietary tool called Plume, our Zeno particle system, and for certain effects, [Side Effects'] Houdini. Because we didn't have many water effects, we simulated them inside Zeno and rendered inside Houdini," says Pearson. In addition, the creature development team handles some simulations. "We do cloth-, flesh-, and spring- based simulations on top of our rigid-body simulations," Pearson says. "That gives them bending and tearing. So their work overlaps what we do. We have a mix of both." For destruction, however, the effects team wanted a new technique that would be faster, better, and easier for the artists. "The big destruction shots are usually the last to finish because they're so complex, so that gave us the lead time to build a new system, refine the pipeline, and solve issues," Pearson says. "Rick Hankins, effects TD and R&D engineer, came on early and probably spent a year in development. He wrote a posi- tion-based dynamics system (PBDyn)." Pearson explains how the new system works: "It takes a set of positions and, based on contact with neighbors, solves constraints. Constraints might include friction, attraction, and repulsion. We also have a shape-matching constraint so groups of particles react like rigid bodies. We can set different strengths for the constraints. The positions are calculated first. It derives velocities from where the points move." Before, the simulation system the team used treated velocity within one continuous volume. When applied in one area, the rest of the grid would compen- sate to reach equilibrium. With PBDyn, the simulation is more akin to spheres in space. As they touch, they exert forces on each other, and the constraints dic- tate how they act when that happens. As a result, one simulation can handle mul- tiple levels of detail and material types. "We start with every particle having the same set of constraints," Pearson says. "Friction determines how far one point can rub against another before it slows down. Repulsion makes sure they all stack on one another. Shape match- ing on a subset of particles moves bigger groups as one unit but with single constraints on individual particles. That way we can have dry dirt on top of a rigid body, and snow on top of that – a set of particles that stick together and operate in the same simulation at the same time." Hankins developed a pipeline around the system so that it can procedurally generate the rocks, snow, and so forth based on models from the model- ing group or the generalist [digital matte-painting and environments] group. A simulation might start with a big chunk of rock populated with particles. Voronoi partitioning breaks the rock into big and small chunks distributed in organic ways. "Then, we manipulate the data to erode certain sections based on edges and distance to the ground surface," Pearson says. "That creates dirt. We defined other areas based on distance from the surface as snow. It's mean- ingless to the simulation – it's just a different setting on the attraction constraint." In practice, an animator might block in the destruction over time – the ground breaking apart, for example. The TDs could use that as a visual reference or, sometimes, as a raw footprint. "In one sequence, the ground is collaps- ing," Pearson says. "Animators roughed that in to show which sections should col- lapse at a particular time. We used those models to drive the sim at first. Then, we decided to use them as a visual guide instead, to make updates easier." For rendering these large simulations, the crew used Clarisse. "Clarisse can handle massive amounts of geometry," Pearson says. "It was great." Barbara Robertson (BarbaraRR@comcast.net) is an award-winning writer and a contributing editor for CGW. the success this team had in achieving their goal. "JJ [Abrams] wanted this to be a film that looked like it could have taken place during the orig- inal trilogies, but he didn't want to ignore modern filmmaking techniques," says Tubach. "The idea was to marry current tech- nology and techniques into the old-school techniques, to main- tain the legacy of the project." Given the studio's history and high level of artistic skills, they were probably the only ones who could. ■ C M Y CM MY CY CMY K