Step By Step: Vantage Point

The Rashomon-styled structure of Columbia Pictures' Vantage Point required Director Pete Travis to depict a tumultuous political rally from the points of view of different characters. Amongst 6,000 people in a plaza in Spain, a tourist played by Forest Whitaker sees one view, while Sigourney Weaver's TV journalist sees another. When chaos erupts and the panicked crowd attempts to flee, the job of depicting these masses from multiple perspectives presented a visual effects challenge for Rainmaker Digital in Vancouver, British Columbia, and the show's London-based supervisor, Paddy Eason.

“The crowds were shot with five cameras simultaneously, widely spaced,” Eason says. “With the few hundred extras that we had, we couldn't address them as we'd usually do: With a one-camera crowd shot where we'd put them all close to camera and let CG deal with the background people.” As a result, Rainmaker would need to generate thousands of digital people using Massive Software, and then intersperse them believably amongst the photographed crowd.

Rainmaker previously had used Massive to populate a stadium in Blades of Glory, but for Vantage Point, the application was different. “These crowds were a lot closer to the camera, and they'd be seen side-by-side with real extras in full daylight,” says Digital Effects Supervisor Geoffrey Hancock.

To help build the necessary modeling assets, Eason supervised a greenscreen shoot of about 100 extras in various poses while on location. “We used a very good digital camera and sent Rainmaker gigabytes of digital stills that they could use to create body shapes and clothes for the Massive people,” Eason says.

Back at Rainmaker, these were blended and combined to produce a variety of looks. “If we had a baggy shirt and a tight-fitting shirt, we could extrapolate an in-between one,” Hancock says. Because some digital extras would be seen fairly close up, Rainmaker created these assets with high-res tools such as Pixologic ZBrush and Autodesk Mudbox and Maya.

Hancock's crew also conducted motion-capture sessions where they recorded a range of behaviors that included stumbling and falling down. “We gathered everything from confident running to crawling on hands and knees,” he says. Rainmaker had licensed a library of Massive behaviors called Mayhem. “That produced some quick results. But we ended up replacing those behaviors because the crowd looked too much like soccer hooligans and not like people who were scared,” Hancock says.

A major challenge was choreographing how the Massive crowd would get plugged up as frightened individuals pushed towards the exits of the walled plaza. This required significant collision avoidance. “Massive has built-in tools to allow one person to see another and choose to walk around them. But when people are packed together, those mechanisms fall apart. People get stuck and are unable to make a decision. We had to create routines that said, ‘If you get this close to the person in front of you, you should use that mo-cap loop — unless there's someone to the left of you,’” Hancock says.

Complicating the process was the fact that Rainmaker had to account for the real people interspersed with the digital ones. That meant creating reference geometry of the extras that the simulated people could react to. Varying the walking or running speeds of the digital individuals was also crucial. “If someone who was running wanted to get past a person who was limping, that added to the complexity of getting people to avoid each other,” Hancock says. “Some of our early sims looked like a choreographed music video, with everybody shuffling at the same speed.”

Once the Massive people had the intelligence to react properly to each other, they were herded toward the exits. “We used sculpted flow fields, which is a technique in Massive where you describe the currents like you would with water. People are inclined to move in the direction that a flow field points them. That was the best way of tweaking this simulation. [The final stage] was to pick out oddball people — like a guy running and waving his arms — and delete them from the sim,” Hancock says.

Rendering the simulated crowd required Rainmaker to do custom programming in order to make Massive communicate with the mental ray renderer from Mental Images. “Massive is a good choreography tool, but it has limited control over the appearance of objects,” Hancock says. “Rather than have Massive decide which person should get what texture map, we decided that at render time.”

Because the original textures had been created with high-res sculpting tools, Hancock had a lot of latitude during rendering. “Depending on how close to camera a person was, more or less polygons were used to create details like wrinkles in their clothing. For a person taking up significant screen space, at render time, more polygons were generated to fill out the shape of the displacement maps that we'd created,” Hancock says.

The simulated crowds were integrated with the plate photography using Eyeon Digital Fusion. “We had an extensive set of render layers, which let us have a huge amount of control during compositing. If something needed relighting, we could do that in the comp,” Hancock says.

Digital Fusion was also used to rotoscope the real people in the plate photography, so that the CG people could be inserted among them.“It wasn't glamorous, but it had to be done so we wouldn't see chattering edges,” Eason says.

Eason supervised all of Rainmaker's work from the production's base in London via daily CineSync sessions. “Rainmaker would FTP 1K QuickTimes, and we'd go back and forth, drawing on frames. My notes would then be inserted into their tracking database. Then we'd do the final approvals at 2K. It only took a couple of iterations at each stage, so it was very efficient,” Eason says. “Of course, I wasn't in Vancouver day-by-day, feeling the pain of getting this done.”

Director: Pete Travis

Visual Effects Supervisor: Paddy Eason

Digital Effects Supervisor: Geoffrey Hancock

Visual Effects Executive Producer: Marianne O'Reilly

Visual Effects Producer (UK): Harriet Donington

Visual Effects Producer (Van): Christopher Anderson

Senior Visual Effects Coordinator: Marta Knapik

Compositing Supervisor (Van): Brian Fisher

CG Supervisor: Sebastian Bilbao

CG Massive Lead: Christina Hsu

CG Mental Ray TD: Alan Hernandez

Pipeline Engineer: Ronald Siy

CG Supervisor (UK): Sally Goldberg