The Mill Gets Massive for the PS2

By Published on .

The latest U.K. spot for Sony's PlayStation 2 represents a crowded milestone in commercials postproduction. The Mill/London, succeeded in creating CG characters with the power of individual thought and action by using a new piece of beta software called Massive, a powerful node-based interface generating characters that appear autonomous within a crowd-characters with individual spontaneity and movement. "Massive generates crowds of people who make unpredictable choices, individuals with mechanical actions that interact with their environment," explains the Mill visual effects artist known simply as Barnsley.

"Mountain," via TBWA/London and director Frank Budgen of Gorgeous, sees civilization clambering skyward with childish innocence to play an epic game of king of the mountain, which is apparently intended as an invitation to join the millions of people who game globally. Production commenced with a six-day shoot in Rio, chosen for its city drops, good weather and diverse casting potential. Set designer Tule Peake (City of God) laid the foundation of the shoot by building a small city on the outskirts of Rio, a metropolitan mise-en-scene for the production crew to film. Six cameras then recorded over 500 extras strategically choreographed climbing buildings and running through city streets with passionate abandon. One exemplary helicopter shot captures 50 stuntmen and acrobats running up a 20-foot cone placed on the top of Brazil's highest skyscraper. Once back at HQ in London, Barnsley applied classic layering techniques to the footage to tweak Budgen's voluminous visuals-filling out empty areas and adding a cityscape to weigh down the spot-before the 3-D department wrestled with Massive.

Mill 3-D artist Jordi Bares stepped in for the larger people scenes, working closely with software developer/Massive creator Stephen Regelous of Weta Digital in New Zealand. Regelous originally designed the software for the battle sequences of Lord of the Rings, but here it was laboriously reworked to evoke climbing, not fighting. Each character, or agent, as they are called, was essentially 'told' which way to go and how to perform by Massive. Specific movements were first motion captured in a studio, then blended into an agent's brain to create programmed spontaneity. "We programmed specific moves and then added over 100 different motion capture moves on top," Barnsley points out. Bares used the information to generate thousands of digital characters, giving each individual a brain to determine how he would react in an environment interacting with other agents. The biggest shot had Bares create 146,000 digital people with unique vision, sound and touch-based artificial intelligence. But all was not exactly smooth sailing: "We had unexpected results programming the software," Barnsley recalls with a laugh, on early simulations and personality traits. "For absolutely no reason, people would jump off the top of the pile or just stand there shaking their hands around like crazy. Then a small contingent of characters just ran away, even though they weren't programmed to do that. The 3-D team had to work it out as they went along."

Most Popular
In this article: