From inputs received at one or more processors, a background audio signal representing background sound is separated from a primary audio signal. The primary audio signal is output with the background audio signal or an altered version thereof according to a user selection between noise cancellation and ambient sound reproduction. More specifically, when the user selection is for noise cancelation, the primary and background audio signals are output with a first altered version of the background audio signal (for example, having inverted phase so as to destructively interfere with the background audio signal); and when the user selection is for ambient sound reproduction, the primary audio signal is output with the background audio signal or a second altered version of the background audio signal (such as a pseudo-acoustic representation of the background sound). One example embodiment is a headset with microphones and speakers for the respective inputs and outputs.