∂Once the room was tracked we used the readymade rigibbody trackers. These are small objects (either bought or made) that have unique configurations of trackers attached to them. When each group of points is selected in Motive it can be locked as a single rigid body object. We tracked a total of three objects and captured a quick motion scene.
In Unreal Engine, we set up the OptiTrack plugin to receive streaming motion capture data and three rigid-body tracking target. There targets are empty actors that will move any movable objects placed inside them in the scene hierarchy, we tried parenting simple geometric objects as well as more dynamic ones such as motion-reactive particle systems.
to stream the tracking data from Motive all we had to do was play back the scene, and we realized that if we go back to capture mode and press the aptly named “Live” button, live tracking data is sent through to Unreal Engine, allowing realtime performance and simulation feedback.
Week 1 afterthoughts
It was surprising to see how easy it is to set up a tracking system using Motive and streaming to Unreal Engine. It got me thinking of the possibilities for performance with live motion capture data and how to reverse the tracking to not place a person in the scene, but map the scene to a person, a-la “Inori”. I hope the semester will provide an opportunity for more physical-interaction based projects using MoCap data and using it to augment a physical space using projection and lights.
Week 2 - Skeletal Tracking
The second week of the lab was focused on capturing skeletal data and mapping human figures in Unreal Engine. This is a more involved process than dealing with rigid bodies on both the Motive side and Unreal Engine side.
We started, again, by calibrating the studio and then went on to setting up the tracking suits. We mapped each actor with a total of 41 trackers in a predefined tracking setup for posture and toe articulation. Once all the trackers were assigned, all we had to do was set the actors in a T-pose, select all of the floating points in Motive and assign a skeleton, just as we do with rigid bodies. The difficulty was in setting up the suits correctly but Motive has some handy realtime visual guides to show you which points are off.
The Unreal scene was already set for us this week, so all we had to do was set the right naming and ids and a blue demon magically started to move around the scene. The downside was that using skeletal tracking requires setting up a control blueprint instead of using the actors, and both can’t be used at the same time, so our attempts to also track some chairs failed for now.
Once both actors were rigged and tracked we had some time to mess around, we used the (surprisingly smooth) office chairs as moving platforms and experimented in recreating scenes of dance, fighting, explosions, flying and swimming. The simulation renders still like bad green screen keying from an 80’s superhero movie but hey, it’s only the second week!
Week 2 afterthoughts
I’ve been learning how to use Unreal Engine blueprints and set up a simple targeting setup to control multiple actors based on motion data. I want to try and create an interface blueprint that will act as an abstraction between Unreal and the tracking hardware, so that the same project could be run with OptiTrack rigs as well as Kinects or Notch trackers.
I’m also working on motion-reactive particle systems that Amit based on distance and speed rather than time. I like the ideas of being able to trace movement without having an actual figure appearing on the screen, on the transient artifacts of motion.
Week 3 - cleanup, retargeting and export
This week we learned how to clean up motion capture data with Motive, use Maya to rig characters we design, retarget captured animations to our rigged models in Motion Builder and bring the joy back to Unreal.
The journey started at the MoCap studio, where Gabriel graciously performed Brazilian numbers, played a drunk and a Gorilla for us. We then set down to clean the data and dealt with unlabeled channels and missing chunks that needed completion. We had some good takes that were done in a matter of minutes, and others that were a total s%!tshow.