House Party ended up taking the most time to capture out of all of SR2’s cutscenes- about one and a half days. I found that breaking up the capture takes into a couple of shots each really helped the actors closely match the timing of the animatic. The longer that you go for a single take, the more out of sync actors tend to get.
The animator responsible for cleaning up the animation data was Will Tam. He did an excellent job of editing the mocap to match the animatic, while inserting some of his own visual flair in. However, not all went well. 3DS Max is known to be a resource hog, and this scene had a large number of characters in it. Oftentimes, our animators would have to hide almost everything in order to get a workable framerate. This, combined with Max’s instability, and sometimes 10 minute load times for scene files, created a difficult journey for anyone involved in this scene, especially Will. It was some combination of Max’s Skin modifier couple with the amount of sheer keyframes in the scene which would bog performance down. Next, facial animation was then hand keyed in after the body animation was completed using morph targets and Volition’s proprietary spline based facial animation system.
After the body and facial animation step, the file is passed back to the cinematic artists. Here, I prepped the scene for export. Mainly this involves setting up a multitude of properties, (who each character is, where this scene needs to play in the world, what level files to load, making sure T-Poses are set correctly, etc). We could also select which characters to export per shot- there’s no sense in exporting and compressing animation for characters that aren’t in the shot. So, our cutscene TA, Nathaniel Albright, wrote a great tool to compute what characters are visible in the shot, and automatically set those properties in the scene. Again, this was another way we that we improved our pipeline over SR1. Imagine doing that for 12 characters across 120 shots for one scene; finding a way to automate this was a lifesaver. Overall, the artists put aside their creativity and put on their technical hats to ensure no property is set incorrectly.
Exporting is long, complicated process; it’s very trial and error based. It’s the slowest and least efficient part of the pipeline. Unfortunately, at this point in production, the audio company we outsourced to needed cutscene renders from in game. They wanted a better sense of the environment where the scene was taking place, and what kind of floor they were walking on so they could make the correct footstep foley, or how big the city was that they were in. Much of that just didn’t show up in our Max preview files.
We had a mad scramble to export scenes into game so that audio could view frame dumped renders of the scene and as a result we had to bring in additional artists and tech artists to help us out. Unfortunately, some of our larger scenes were planned without much concern for memory issues, or other technical issues (too many characters in a shot, characters too far away from camera so they don’t render, too many locations loaded into memory… etc). We wanted SR2’s cutscenes to be bigger and better, and they certainly were, but that proved costly at the expense of time and manpower. Once a scene was relatively stable, we could render out a frame dump for audio and move onto lighting.

