I'm afraid you can pretty much forget all that because the video was shot the wrong way in the first place. The camera is far too close in and the artist obscuring too much of the background to get a stable track. His gestures would throw off the tracker, also. Aside from that, due to how it moves, you would have to do it for every shot separately and still need to manually matchmove sections where the motion is simply too fast. Additionally, you cannot go creative with contrast and color. This needs to be shot as neutral as possible. In some of your footage there is nothing but black shadows and while that may be due to the color correction, I doubt the original sources look that much better. so ultimately, you are stuck with those techniques for static shots and throwing a lot of manual work on top of it to make it work in the few shots where it may be possible to kinda match the camera..
Well I will be cutting away to other shots between the few parts I want to add 3d text/objects to. The actual shot is alot better in terms of exposure but also he is very upclose to the camera as well...perhaps I will upload the actual shot I'm dealing with tonight and get some feedback.
How difficult (I don't mind putting in the work) would it be to correct manually?...I'm maybe talking about 2-3 seconds per graphic appearing then disappearing.
Could I possibly track his eyes/face and lock to that...masking around him and anchoring the graphics to the artist rather than the surroudings???
What do you think?