Seth MacFarlane, writer/director and voice star of summer comedy Ted 2, recently relied on Xsens‘ lightweight MVN inertial motion-capture suits to capture his performance as the titular Teddy bear in the film, which follows Ted’s attempt to persuade the Commonwealth of Massachusetts that he’s legally a person.
“Optical mocap wouldn’t have been feasible given the demands of the production,” said postvis supervisor Webster Colcord. “We needed a portable system that could be quickly set up anywhere with minimal impact on the rest of the crew.”
Duplicating the original film’s capture process, Colcord and his team used a MVN system to record MacFarlane as he performed the character live. The sensors were strapped over his normal clothing, leaving the director free to carry out his other duties.
For the big musical production, which sees Ted accompanying a large group of dancers, mocap was recorded with three separate professional dancers and combined to create Ted’s impressive moves. To pull this all together, Colcord used MVN Link – Xsens’ next-generation lycra suit, which enables users to capture the motion of actors at up to 240Hz, and at distances of up to 150 m.
“The dancers could run as far as they wanted without us ever having to constrain their distance,” said Colcord. “Not having to worry about staying within a volume [as with optical motion capture] is a great advantage, as is the fact that we don’t have to worry about markers being occluded.”
While accurately recording the motion of a dancer with both feet off the floor would previously have been the sole preserve of optical capture, Xsens’ recent advances in inertial capture technology made even high-intensity sequences possible. “We had three very active dancers flipping, jumping, spinning, and more,” said Colcord. “It was really fun to deal with the challenges of that situation – and a great test of the new hardware and software system.”
The team captured three streams of data, all synchronized by timecode – video for performance reference, the actual motion-capture data and a recording off the video card of the live mocap streaming into Autodesk‘s MotionBuilder software, where it was retargeted onto the CG character.
“Xsens’ tools for streaming to MotionBuilder and outputting data in various configurations are simple and easy to use,” said Colcord. “We have an in-house postvis rig that we built for Ted, which is very lightweight and has a multilayered skeleton, which allows us to offset the mocap [in order to adjust the performance] without animation layers, and without being destructive to the original data.”
For the postvis – a rough version of the visual effects, used to define the look and timing of shots early in the creative process – the team used a set of standard software packages, including SynthEyes and MatchMover for camera tracking, Maya for animation work, and After Effects for compositing.
“After a typical [recording] session, editorial cut the video elements in as picture-in-picture with timecode burns,” Colcord explained. “I’ve developed my own method for syncing up the data in Maya, based on timecode, and that has been pretty successful.”
When the process was complete, the assets were passed to VFX facilities Iloura and Tippett Studio to create the final effects for the film, using the postvis as reference. “The synced data was passed along with minimal corrections,” said Colcord. “The VFX artists baked it onto their animation rig through a constraint setup and did offsets on animation layers.”
Thanks to the freedom afforded by Xsens’ MVN wearable motion-capture systems, MacFarlane was able to ‘be’ Ted live on set, enabling the other actors to react realistically to the bear, with the confidence that the final CG version would match every nuance of his performance.