Hoping someone can direct me if there's a better, more efficient direction I should be going in. I've been motion tracking to blur a face. It's someone that is doing a tutorial and will sometimes look at the camera and sometimes turn to look at the board (placed at 90 degrees to the camera), so that the face goes from looking dead on at the camera to being in profile.
I placed a blur that's large enough to cover the face when he's head on to the camera. But when he turns to the profile, I feel that the blur is a little big and covers too much of the background in one direction in addition to covering the face ... so it looks like a lazy/sloppy job to me. For a polished production, I would expect the blur to be tight enough on the face at all points to just cover the face. (The face, obviously, takes up lesser screen real estate/space when it's in profile than when it's facing head on towards the camera).
I tried including the rotation and scale in the motion track analysis, and that seems to work decent enough, but of course I still have to spend a lot of time cleaning it up ... though I like the scaling that it's doing, and the result it's giving me is what I sort of have in mind. If this is best way to do, then it's all good. But thought I'd ask in case anyone has any pointers to better techniques.
Most "face blurring" jobs I did required just a simple circular mask to track the face and a few seconds to play with the feather on the edge of the blur. If for some reason you need a tight blur on the face, I would see this as a job for Mocha. There will be roto involved, but I don't see how you could get away from it.