Disclosed are systems and methods for augmented reality multi-view telepresence. An exemplary embodiment is provided for a first-location participant and a second-location participant, each location having respective pluralities of cameras and display segments. A session geometry of a multi-location telepresence session is obtained for the participants. According to the session geometry, a selection is made of a first-to-second-viewpoint second-location camera from among the second-location cameras and a first-to-second-viewpoint first-location display segment from among the plurality of first-location display segments. A first-to-second-viewpoint video stream captured by the selected first-to-second-viewpoint second-location camera is received, and a line-of-sight augmented-reality session is generated by rendering the received first-to-second-viewpoint video stream via the selected first-to-second-view-point first-location display segment.