Hey Richard
Yes, we may have had some trouble in explanations/understandings, however as a good piece of news, I think I understood what your goal is.
Provided you cam is moving (because if it's not, it could be easier)Look, if "this content marked in blue" was in 2D only, which it is if you pipe anything coming from
after the Rendere3D node, then you should put a tracker node > track relevant and idealistic points representing the 4 points of the desired target > turn the tracker node mode into "corner positioning" (and pipe your guitar player in it for the tracker node then acts lime a Merge one) ...
Of course, if there are too large PTZR or even just huge circular moves, it will be distorted, and in any way, the content should obviously be filmed/made at the same position angle in the first place.
Now, what I wrote is about 3D, which is the right way to do it. Then, the 3D content must be part of the comp - before the Rendere3d - and be placed where the results of a CameraTracker solve tells you to. This is now obvious that you content is a real 3D one
(if not, you can add a Shape3D after it to be able to include it but in case of large cam moves, it doesn't sell anymore)So in order to be specific - and not write down the whole process which would be a novel here
- here are a couple of videos showing you what to do with it. Some nodes must always be there (merge3d, camera3d, etc) but the CameraTracker solve quality is your main goal...
It depends wether you have time to watch a long one, or shorter ones