Hi all !
 
 
I need to track an animals facial movements. In a test with a locked down camera and a dog I had no problems att all to track and replace the face of the dog. On the shoot however, camera movement will be hard to avoid so I need to test this to.  
 
This is how i did my test:
 
The camera is on  a tripod and paning / tilting slightly to frame the fox's movements. I apply the nodal pan constraint, mask out the dog and do an automatic track of the camera movements. Then I add som manual tracks (8 - 10) on the fox´s face and group them and aply mobile to that group. This produces a sollution where all the locators from the background are projected inside a "sphere", as expected with a nodal. The locators from the fox´s face however are spread over a very unatural depth. How can I avoid this ? What kind of info could I supply MM with in order to get a more accurate solve ?
 
 
Thanks In advance /
 
 
Martin Malm

First, you must be sure that the dog face can be considered as a rigid-object, else you'll have to use motion capture with multiple camera.
 
Then, is the depth inaccurate because of a lack of parallax ?
Or, if you prefer, have the dog face track a good residual, or are they considered bad by the solver ?
 
Did you really need to track both the nodal background and the moving face at the same time ?  
 
If you can you send your data to support@realviz.com for a deeper analysis, it would help.