A telepresence device is disclosed, which includes a display and a camera array including cameras configured to capture images of a local subject through the display and output a plurality of local captured images. The telepresence device also includes a local tile selector configured to select a plurality of tiles from the plurality of local captured images based on remote subject point of view (POV) data received from a remote device. The local tile selector outputs local image data that comprises the plurality of selected tiles. The local image data is sent to the remote device to composite an image of the local subject.