There are people doing mocap with the Kinect sensors. They’re not free, but they’re a lot cheaper than just about any other form of mocap. Here’s a video someone did. Near the end they have problems with occlusion because they only had one sensor. Having two or three sensors helps eliminate this.
Learning to keyframe is free. Give it a try. There are lots of tutorials on Youtube.
Are you learning to use the curves editor? That’s how you finetune your animation.
Is using pre-recorded footage MOCAP? Is it easier/cheaper to do this?
Yeah, most of those files will be mocap. The problem is that you’re limited to what someone else has done, and lot’s of other people will also be using them so there’s nothing unique about your animation. Taking the time to do it yourself will be worth it.
Some people are leveraging videos of themselves for automatic animation. It is a different approach, however, and automatic might not be the correct term. Check out Sebastian’s work with head tracking.
I have also seen some work on tracking dots on face to handle expressions on 3D characters.