Back when I was with the Box at Rhythm and Hues, there was a time when we didn't have access to motion capture, so when we needed extra realistic human motions, we relied on a system we called 'rotocapture'. The idea was simple--record the performance on video from at least two angles, save out the frames and map them onto polygons in the 3D application. Then, using the images on the polygons as reference, we matched the character poses to what was in the video.
The results was something that was similar to but not exactly the actor's performance, because the process was also subject to the animator's interpretation of the motion.
How's that for muddying the waters?
As for mocap vs. keyframe, I think it depends on the intent of the project. Mocap doesn't necessarily save you time and money for every production, but it will certainly give you different results. When we did 'Mercedes Beast', my boss wanted to use mocap for the monster but the artists (myself included) on that production argued that it wouldn't look right because it wouldn't be 'beastly' enough. After a few tests, he agreed. You can see some of the results on
my demo reel. Then, a couple of years later, we did a series of 'Call of Duty' commercials, and this time we fully supported his decision to use motion capture--we were a tiny crew of artists in the Box, and none of us saw any advantage in animating dozen's of human soldiers going through fairly 'normal' human motions, especially since we had our hands full already, building the environments, vehicles, and creating the visual fx for the projects. And even though we had motion capture on that project, there was still a fair amount of manual keyframe animation, with or without the mocap as a base. In other words, while we often found motion capture to be a wonderful tool for many of our video game cinematic productions, it was not always good enough to serve as final animation by itself.
So, it's really a matter of choosing the right tool for the job.
G.