Audio in the starter stage for third person seems to be based on avatar location and orientation. I'm sure this is a temporary idea, but I'd like to offer my suggestion that I think the best balanced way to implement this in the final version is to keep the audio sounding as it does from the avatar's head, but track the based on the camera.

I was spinning the avatar around and it was quite silly hearing the audio change as if I was spinning, when my camera was not. I thought maybe attaching the audio listener to the camera position orientation would make sense at first, but honestly I think that has a higher chance of breaking worlds and immersion if desktop uses can cheat the audio listener location by going into third person like that. I think binding the stereo balance to the camera rotation, while keeping the position locked to the avatar head, is going to result in the smoothest experience across all forms of movement in and out of third person.