A method and apparatus for controlling a live avatar include features for directing the live avatar gaze to be consistent with a remote operator's view of the live avatar's surroundings. Indication of gaze direction may be continuously provided to the live avatar using movable optics, for example a heads-up display, or a movable optical prism or lens. Gaze direction or operator commands may be communicated using a haptic device. Video and control streams from multiple live avatars may be aggregated through a host server and allocated via an online interface to multiple prospective remote operators.