Multimedia
Blogs
Forum
Affordable HD
Whitepapers
Newsletters
Events
Advertisers
DCP Directory
MillimeterAnimated Performance
Nov 1, 2007 12:00 PM, By Ellen Wolff
Sony Pictures Imageworks goes for a more live-action style in motion capture.
Refinements in muscle-simulation technology were crucial to the animation of a believable warrior.
For the actors, whose performances were captured on a 25'×25' stage, the mo-cap recording felt like a stageplay. “I heard both Anthony Hopkins and Angelina Jolie mention that they didn't have to worry about playing to a particular camera angle, but could focus on the other actors. It was shot in full scenes, and they went very fast,” says Animation Supervisor Kenn McDonald.
A key advance that Imageworks made with Beowulf was the simultaneous capture of facial performances. Each actor had an HD camera dedicated to getting close-up facial-reference footage, but tracking markers also captured the subtleties of face and even eye movements. “We found a technology used by ophthalmologists called EOG — electric ocular graphography. Electrodes measure the electrical impulses that move the eyes, and we had them wired into mini-pocket computers. This data contained almost subliminal nuances, and we captured that at the same time we were capturing the performances of the actors' faces and bodies, so we could sync them up,” McDonald says.
The rigs used for facial capture were also customized for each actor's performance. “Some actors could raise their brow higher than others, and their cheeks bunched differently. Every facial rig ended up being specific to that actor,” McDonald says. “So if Anthony Hopkins comes back, we're ready to go!”
“On Monster House, we had the introduction of a pose-based system called FACS [Facial Action Coding System],” McDonald says. “First you have an actor go through a series of expressions and poses and phonemes, and then you recreate those expressions and phonemes on the digital character using a facial rig. You take the performance-capture data for an actor, and using a matrix of all these expressions, you do a solve that basically breaks down any given frame to a combination of expressions and phonemes. That would give us an initial impression of each performance. Then the animator, using a set of controls laid over the top of that performance-capture data, could finesse the expressions to make sure we're nailing the intent of an actor's performance.”
“I don't like to use the word ‘realistic’ because obviously it's not real,” McDonald says. “We're not trying to convince people that's actually Anthony Hopkins — it's a version of him. We tried to capture an authentic performance that's accurate to his original intent, with a high level of detail. If Anthony had only recorded a voice and we were crafting the visual performance based on a limited amount of video reference taken during a voice session, we'd have 20 animators trying to maintain a consistent performance. By having a single actor drive the performance, the little twitches he used to create his character will be consistent throughout.”
“We've gotten better at sorting through the tracking dots and finding the nuggets that we need to apply to a character. But everything you see in Beowulf is an actor's performance. Mo-cap movies are actor-centric, and they'll continue to be,” Chen says.
Once the performance data was tracked in MotionBuilder, animation began in Autodesk Maya. Imageworks also applied its newly refined muscle simulation system. “Beowulf likes to take his clothes off, and he runs around naked for a lot of the movie. Our system allowed us to simulate muscle jiggling under skin. It required an iterative process to achieve the authenticity that we were shooting for,” McDonald says.
