I did an awful lot of hand keyed animation during my time at Irrational Games, but the use of motion capture was always a mainstay in the pipeline. Mocap was a part of our process and we were expected to use it when we could, but we would always approach using it from the perspective of what was considered best for the scene. It was never used in the File>Import = done context. On the contrary, the need for serious iteration was almost always necessary.
I should mention, before I go any further, that the actors and actresses that worked for us did a fantastic job. What I’m about to talk about by no means belittles what they did for the project. It was their job to act out the characters in the Bioshock universe and they delivered brilliantly. Conversely, it was our job as animators to ensure that the scene read clearly and best represented their efforts. This usually meant a lot of iteration on our part to make the scene work. In my professional opinion, the biggest problem with using motion capture for a cinematic scene is that the amount of iteration needed can sometimes take longer to edit than if it were hand keyed in the first place. In some instances we had to simply start over because there were times when there was actually too much movement (again not faulting the actors and actresses). In these situations you would lose clarity because of the relative motion between character and almost motionless camera. The closer a character was to the camera, the problem was more apparent. Personally, I would consider the mocap data as reference first and foremost. I would use what worked, but I was never afraid to take liberties for the sake of the performance and the scene as a whole.
I’m going to use the dance scene on the docks of Boardwalk to show my process. Here is Elizabeth’s performance actress, Heather Gordon, performing to the dialogue during the capture session.
Here is the raw motion capture data from the session imported into 3D Studio Max. From this point, the direction I was given was to push Elizabeth’s performance to be excited, innocent, naive.
Motion capture data stamps a key on every frame – making editing a nightmare. Some animators like to use layers to edit the base motion from this point, which is fine but I’ve found layering to easily cause confusion even if you’re really organized. Layers can also cause an animator to forfeit control in certain situations. Personally, I prefer to have complete control of one layer in the scene so what I’m seeing on the timeline is what is really happening. So I bake the keys on every other frame to make editing a little more manageable. After baking on 2’s, I would start to take a look at the timing. Without going into the graph editor, I start deleting and shifting keys to make a hold cleaner, a transition faster or to cut something that seems “over-animated” (too much movement). I then start thinking about the poses themselves and push them for maximum appeal and readability. In some cases, I’ll add acting beats for what I feel the subtext is in the dialogue. Here is what it looks like after I’ve done my blocking pass with the motion data.
Then I polish and making sure it plays cleanly in the game. My personal goal was to push her appeal to the point where the player would fall in love with her at this moment. :)