One thing which I did find was that it doesn't always get where the posed character needs to be in relation to the depth. The wiki says "don't worry about it, just delete the pose and try again" but the wiki wasn't posted when I initially played and what I figured out that worked really well, was to set the character size like we do with multiple cameras, then place it into the frame left and right and vertically as normal but then...

...right click the character and move the view point to see the scene from the side, you'll then see where in the depth the actor appears and then just move the character to the depth position of the actor.

Using this method, I never failed to get an instant fix for the chatacter to actor initial pose.

I really like this, not sure which I like best, the excuse to buy a Kinect or to have fun with the software.

I'm very happy with the "lack of" real time. If the accuracy of the tracking, which is already very fast (I was getting 0.4s per frame yesterday) means that there's less error in cleaning up real time track with errors then it's a result all round.

Two words left... "head tracking" :)