The American Society of Cinematographers

Loyalty • Progress • Artistry
Return to Table of Contents
Return to Table of Contents April 2006 Return to Table of Contents
Inside Man
Sundance 2006
DVD Playback
Post Focus
Djarum Mezzo
ASC Close-Up
Guava Performs Massive Feat

To drum up support for community-service programs, United Way created the “Lend a Hand” campaign, a series of PSAs starring players in the National Football League (NFL). One of the spots focuses on children’s literacy and features Atlanta Falcons offensive lineman Keith Brooking, as well as more than 100,000 digitally generated children. Created by New York-based Guava in collaboration with freelance artist Philipp Hartmann, the effects were done largely in Massive Software, an application that generates 3-D crowds in a way that is different from most computer-animation programs.

The PSA opens with Brooking reading Dr. Seuss’ Green Eggs and Ham to a small group of children. As he holds up the book to show them the pictures, the view widens to reveal he is sitting in the middle of the field at the Los Angeles Coliseum, reading to a stadium full of kids. As Brooking takes in the magnitude of the task before him, the voiceover announcer appeals to viewers to help by volunteering for literacy programs through United Way.

Alex Catchpoole, Guava’s visual-effects supervisor on the spot, says everyone knew from the start that the stadium crowd would have to be done as an effect, but they weren’t quite sure how to accomplish it. The traditional way to approach the task would have been to shoot a number of plates, moving extras from seat to seat to fill the stadium in stages, and then composite all the different views in post. The problem with that was the scale of the Coliseum, which seats as many as 100,000 spectators. “We quickly realized that unless we could have thousands of children there on the day, we would be there for months trying to fill every seat,” says Catchpoole.

The solution turned out to be a combination of 150 real children and more than 100,000 digital kids created in Massive. “The children closest to the camera were all real, and as you get farther away the digital kids begin to take over, intermingling with the real kids,” says Catchpoole.

After the stadium material was shot, the Guava team began by filling up the rows immediately in front of the camera with footage of real children to create what Catchpoole calls a “real-kid buffer” between the viewer and the digital kids. “That turned out to be a little more complicated than we expected because the kids fidgeted, and some turned to the camera and waved,” he says. The artists ended up doing “quite a bit of head replacement” to make the real children look the right way.

Massive’s digital characters needed something to sit on, so Hartmann created a replica of the stadium and its seats in Cinema 4d. “We got in touch with an architectural firm that had worked on the Coliseum, and they sent us a blueprint of the stadium,” says Catchpoole. Ironically, that digital model is never rendered as a visible element. “It’s kind of a ghost that the digital kids could sit on so that when they were combined with the live-action footage of the stadium, their positions would match.”

The model was then handed off to Hartmann so he could create and animate the crowds. Hartmann notes that whereas most animation programs are based on key framing — wherein the artist sets certain poses for characters and the computer fills in the frames in between — Massive is based on artificial-life technology, which means that characters are created, assigned certain traits, and then allowed to respond to stimuli. “You have to rethink the whole process,” says Hartmann. “For the first two weeks, you’re looking for the buttons you’re used to, but Massive’s method is actually much easier.”

The program ships with a number of built-in characters called “Ready to Run Agents,” which can be modified to suit a given project. In this case, Hartmann started with a character called Stadium Guy, who performs actions common among spectators at big sporting events — clapping, cheering, standing up to see what’s going on, for example. Hartmann worked with two other freelance artists, one in Israel, the other in San Francisco; they used photos taken of kids on the shoot day to create new bodies that could be applied to Massive’s Stadium Guy skeleton. The whole figure was scaled down to child-like proportions, and skin color, hair and clothing were mapped onto the figures. “We chose from different sets of variables, and Massive multiplied those so we ultimately got thousands of variations of that one kid,” says Hartmann.

The new character, dubbed Stadium Kid, was then given some new actions specific to the ad’s script, such as straining to hear and fidgeting. “By setting variables, we could control how active they were, how likely they were to scratch their heads or shuffle around,” says Hartmann. “During the shoot, I thought the kids were crazy and difficult to handle. But in the CG realm, we had to calm down the digital kids’ motion so they’d match the real kids.”

Catchpoole and Hartmann found that the amount the digital children moved had to be scaled up or town, depending on where a character was sitting. “”There was quite a bit of direction involved, similar to extracting a performance,” says Catchpoole. He explains that a level of movement that worked well in the foreground would be much too small and hard to read in the background — “It looked like everyone was asleep.” He adds, “But if we’d had 150,000 real children on the day, the same issue would have come up — and a lot of others would have come up too!”

Once the actions began to look right, Hartmann started rendering out the crowd in sections, about 7,000-8,000 children at a time. “Each frame took about five minutes to render, which wasn’t too bad considering the amount of geometry,” says Hartmann. Guava artists then composited the sections with the original footage in Flame. Catchpoole also did a little extra finessing in Flame. For one shot that was relatively close to the kids and looking across the backs of their heads, the digital children weren’t mixing that well with the real ones. Using stills from the shoot, he went through the shot and did head replacements, tracking still images of real heads onto the backs of the digital children. “About 10 rows of digital kids have real heads, which really helps sell the scene,” he says.

Hartmann says Massive’s artificial-life technology proved perfect for the United Way spot. “It’s literally dynamic, which makes it so realistic,” he says. “It imitates life in a way you don’t control. You just set up an environment and let it go.”

Related Links

<< previous || next >>