L.A. Noire - Rockstar
The quest to improve videogame
animation is never-ending. Twenty years
ago, Jordan Mechner created rotoscoped graphics for the original Prince of
Persia by studying films of his little brother running and jumping. Half a decade ago, Half-Life 2 wowed us with
facial animations that made use of research by behavioral scientist Dr. Paul Ekman.
Today, the bar's about to be raised again, thanks to the creation of MotionScan, a new system created by Sydney-based tech company Depth Analysis, and featured in Rockstar's upcoming period crime epic L.A. Noire.
The goal of MotionScan is to remove, as much as possible, the intermediary of a human animator between the motion-capture and the final result. The
actor's every facial expression must be captured in such precise detail that a
computer algorithm can spit out usable CG imagery without any need for
Aaron Staton in the array How do you do that? By setting up 32 precisely-positioned cameras
in a sophisticated array. The actor sits
in the middle, reads her lines, and via some fancy geometry, computers are able
to reconstruct a complete 3D rendering of the performance. From there, it's just a hop, skip, and a jump to place these fully-rendered graphics in the context of a videogame cutscene
or gameplay sequence.
Brendan McNamara, who oversees Production and Technical Innovation at Depth Analysis, boasts that the resulting fidelity is so good, it exceeds even what
James Cameron was able to achieve with those fancy head-rigs he used to shoot
the motion-capture for "Avatar."
We visited Depth Analysis's Culver City studio, where the copious motion-capture footage for L.A. Noire — the game's script clocks in at over 2,200 pages — was shot. Over 400 actors performed, done up Hollywood-style with period hair and makeup, with lead actor Aaron Staton (known to TV fans as Mad Men's Ken Cosgrove) alone logging 50 days of
Sitting in the middle of the camera array, actors read their lines while looking at a taped-up picture of the "Mona Lisa" to ensure their eyeline's going in the right direction. A massive (and noisy) hard drive array stores the terabytes of video footage generated, and a bank of beefy CPUs crunches the necessary numbers to spit out usable 3D imagery.
Real life and MotionScan. How does the result look? Impressive, to say the least. We watched some of L.A. Noire in action, and the facial expressions are eerily convincing.
In a title where interrogation of untrustworthy suspects comprises a key component of gameplay, it's essential to have virtual performances in which every flicker of the eyes can be observed — and the game lives up to this
challenge. The only occasional awkwardness comes in the mixture of the ultra-realistic facial animations with the more-conventionally-captured body animations, which seem a tad clumsy in comparison.
McNamara envisions big things to come for MotionScan, and not only in games; he sees the tech coming to the movie world, too. When complete performances can be captured from every angle, we're that much closer to having sophisticated entertainment that truly lives in the third dimension.
You can see MotionScan in action
in the L.A. Noire trailer here.