Couldn’t you also have a separate model for each? Rather than splitting the head and creating double the draw calls for the character, have a model that doesn’t have the head, and switch out during the camera movement. Could be tricky to do correctly, but would be easier, wouldn’t it? After all, in that view, what you would normally see is the hands and any objects. So you can create a much smaller mesh with just those objects.
But, you could also make the head see through, by switching between two materials as well, though this would also increase load.