Like its misanthropic superhero, director Peter Berg didn't mean to create havoc off the set of Hancock (now on DVD and Blu-Ray). It's just that his run-and-gun style left the film's visual effects department, Sony Pictures Imageworks—accustomed to longer lead times and meticulous advance planning—scrambling to keep up. But Digital Effects Supervisor Ken Hahn is proud to report that the team rose to the occasion and even embraced opportunities to impress its fearless leader, beginning with one jaw-dropping shot affectionately dubbed "butt-head."
In the midst of a public relations campaign for redemption, the film's eponymous lead character, played by Will Smith, is thrown in the slammer with every criminal he's ever manhandled. And now they're eager to manhandle him. Warned by Hancock that if the harassment continues, he'll shove one inmate's head up another inmate's, um, bum, they persist and Hancock can't resist being true to his word.
"When I read that scene in the script, I laughed out loud," says Hahn. "I assumed, like everyone else in pre-production, that it would be sleight of hand, you'd never see the gag on-screen." So, that's the way the live action plate was shot, from several angles with multiple cameras. And Hahn thought the rough assembly of the scene was pretty funny. But, according to Berg, one thing was missing: the "money shot" of the offending inmate's head stuck in the other inmate's backside. This is the part where jaws dropped.
"We had already wrapped principal photography and, because we didn't anticipate this as a visual effects scene, we hadn't acquired the necessary data," Hahn recalls. That data would have included full body scans of the two actors involved to create digital doubles, texture photos for hair, skin tone, and eye color, and digital stills of the set. To his surprise, far from daunted, Ken's team was actually excited about the shot. "They said, 'We live for stuff like this!'" he laughs.
Their objective was to create a composited live-action and CG wide shot of the two actors, visible from head to toe, in the correct position and environment, and properly lit, while the other actors, including Will Smith, react to the gag. Ordinarily, the team would start with a reference shot, but "we didn't have it, though we did have alternate takes of the shot from multiple cameras, all with similar lighting conditions." This data was used to re-construct the environment and begin to build the CG characters.
"The inmate whose head is in the, uh, unfortunate position was entirely CG in the missing shot," says Hahn, "and the other inmate was partially CG." The team wanted to preserve the head and chest of the partial CG character because "we don't want to try and replicate an actor's work," but they did add a pair of virtual arms, flailing against the invading force. According to Hahn, it took four team members four months, working on "lots of cloth simulation with Maya, really working on the texture aspects" to get it right. "One person worked on muscle rigging [layering virtual muscle with human properties on a CG skeleton]," Hahn recalls, "one person paint-fixed the entire plate [to remove unwanted objects from the frame] and worked on plate re-construction, a third person worked on cloth simulation, and a fourth worked on the lighting and composting." In the end (ba dump bah!), Hahn acknowledges that it was worth it. "That scene gets the biggest laugh in the movie."
If the VFX team were caught off guard with the prison sequence, they were almost over-prepared for Hancock's inadvertent train derailment scene. "We had pre-vized the whole thing [created a rough, animated comp in advance of the shoot]," Hahn recounts, "ending on this beautiful dolly boom shot [the camera simultaneously moves back and up] that reveals all the destruction Hancock has wrought by trying to save the day." But the problems with pre-viz, as Hahn sees it, are two-fold: first, filmmakers get wedded to virtual camera moves that often can't be replicated in the real world, even in this era of computerized rigs and cranes. "And," adds Hahn, "we spend so much time inputting data into these animatics—and it's inaccurate data."
For instance, when the production team got to the actual train yard location in Long Beach, California, it was different from its pre-visualized counterpart. "The biggest difference was the unachievable camera move due to the differences between the pre-viz model and the actual location" says Hahn. The camera's choreography couldn't be achieved with the techno-crane (a computerized rig that controls the camera's movement), and its resting position was not high enough to take in the whole tableau of destruction: Hancock, the train wreck, all the collateral damage, and lots and lots of gawkers.
"We were all frustrated that the scene was not working out the way we had anticipated," says Hahn. "We started with Will Smith, the first train car, and the railroad tracks, plus several cars in the [live-action] plate, then inserted additional CG train cars and automobiles, replaced a bridge in the background, removed a palm tree in the foreground, layered in the Los Angeles skyline, and added lots of debris and other destruction." Still, using the pre-visualized, computer-generated camera move was not resulting in the epic reveal the filmmakers were after. "We finally realized that the reason it wasn't working is that we were slaving to the original animatic," says Hahn. "As soon as we ditched the animatic and approached the scene from a fresh perspective, everything fell into place, which was very gratifying."
Pages: 1, 2