You could disable all the rendering code in OVRCamera and OVRCameraController, disable both of the eye Cameras, and then add a Camera to a GameObject that's half-way between the eyes. You could achieve that by making it a child of the left eye and making its local position (0.5 * IPD, 0, 0).
Hey vrdaveb, I have tried to disable the rendering in code but I have not been successful. I couldn't figure out how to deactivate the render without turning the tracker off as well. Could you explain a bit more in detail how to switch the render off? That would help me tremendously with my current project.
@FisherDachs: Does the person wearing the rift need to be able to see the environment? In your demo video it looks like the person would need to see the environment in order to make a "convincing" recording. If so, I have two solutions that may help you.
1) You could record the player's head movements and play them back and capture that as your undistored first person view. 2) You could network your VR experience with the Rift as the server and the undistorted and webcam view as the client. If you attach a networkview to the OVRCameraController and sync the ID with a camera in your client you should be able to run both players at the same time and use some video capture software on the client window.
Personally I think option #2 is way easier to implement.
EDIT: Latency for the network solution really shouldn't be a problem since the server is localhost.