Assisted viewing of an indexed audio video stream is accomplished by enabling natural language command of an audio video stream repository to control the positioning of or selection within, manner of playback from, or other characteristic or action related to the viewed stream. Embodiments process spoken or textual commands, couched in ordinary natural language, which may include descriptive prose, or references to named persons or entities depicted in the stream. Both pre-recorded and live action audio video streams can be processed. Such processing can include creation anew or augmentation of an associated audio video stream index. Numerous particular useful commands, especially those pertinent to audio video streams of sports contests, are described. Embodiments may set alerts that are triggered by notable events within live action audio video streams, with notification of the user when they occur.