Abstract: Techniques for managing sleep states by selecting and presenting audio content are described. Disclosed are techniques for receiving data representing a sleep state, selecting a portion of audio content from a plurality of portions of audio content as a function of the sleep state, and causing presentation of an audio signal comprising the portion of audio content at a speaker. Audio content may be selected based on sleep states, such as sleep preparation, being asleep or sleeping, wakefulness, and the like. Audio content may be selected to facilitate sleep onset, sleep continuity, sleep awakening, and the like. Audio content may include white noise, noise cancellation, stating a user's name, presenting another message such as a recommendation, the news, or a user's schedule, a song or piece of music, and the like, and may be stored as a file, generated dynamically, and the like.