LA Noire secrets revealed

T3 chats to Rockstar Games about the detective epic

Showcasing some of the most sophisticated motion capture technology seen in a video game, Rockstar and Depth Analysis reveal how one of the hottest games of the year came to life

- LA Noire review: Get the T3 verdict

Reading body language is a key aspect of LA Noire. Was this part of the original plan, or was it born out of what the
tech made possible?

Rob Nelson, art director at Rockstar: “The MotionScan technology was developed in parallel with the game, and they both evolved over time. “[Developer] Team Bondi had been exploring new ways to capture the subtleties of a person’s
facial expressions – the ‘outside’ of a person rather than the skeleton movements of traditional motion-capture data. “The ability to use the non-verbal cues and emotions as guides during interrogations was something entirely new, and added a layer of depth and consequence to interactions.”

What was the biggest logistical challenge?

Oliver Bao, head of R&D, Depth Analysis: “MotionScan captures a massive amount of the detail of a performance in the form of data, so the main challenge was figuring out a way to bring that into a video game when the rawcapture data is 1GB per second; 21 seconds of capture would fill up one layer of a Blu-ray disc. “The compression technology we developed allows each character’s face to run at about 300k per second and still look very good. This was a huge achievement considering LA Noire contains about 21 hours of face-capture content.”

How did you go about pushing the boundaries?

OB: “Our approach was to go outside the box and try to generate 3D data from traditional 2D video. It’s not motioncapture – we went the opposite direction by trying to represent the 3D surface, instead of confining facial movements to the bones. We are firmly of the view that an animator cannot animate an actor’s face better than or equivalent to the performance that the actor can deliver. It’s pretty fundamental.”

How does it work?

OB: “An actor typically goes into hair and makeup as in film or TV. Once they’re ready to go into the MotionScan rig, we mic them up and load all of the lines for the scenes. We tend to keep the sessions to an hour as it’s pretty isolated being in a soundproof room. “The director is in an adjoining room and has a close-up camera view of the actor. We generally start shooting with small sequences that can be used for cutaways when the actor isn’t speaking but we want a reaction. These include emotions: happy, angry, sad, proud, humble, etc. We do about 15 of them.

“When a session is finished, we archive the data for backup and then send it to Sydney for processing. The final output is then checked for quality, and Team Bondi’s character artists, designers and animators import the data into scenes and locations for review in-game.”

Is building the studio set-up a lengthy process?

OB: “It took a month of 16- to 20-hour days to prepare. There are more than 1,000 pieces to the capture rig alone, and it requires a suitably sized soundproof room with quiet air conditioning and a large amount of power for the servers that capture the data. Then there’s calibration, functional testing and stress testing to do…”

So where next with this?

OB: “Part of our road map is to do full-body capture so we can get the complete actor, head to toe, in costume while they perform. It’s incredibly challenging, but there is a lot of buzz from investors for that kind of tech

You can check out the official LA Noire video below to see how it all turned out

T3 Video: LA Noire: The story so far

Source: T3 Video