Various embodiments of the invention provide a facility for monitoring audio events on a computer, including without limitation voice conversations (which often are carried on a digital transport platform, such as VoIP and/or other technologies). In a set of embodiments, a system intercepts the audio streams that flow into and out of an application program on a monitored client computer, for instance by inserting an audio stream capture program between a monitored application and the function calls in the audio driver libraries used by the application program to handle the audio streams. In some cases, this intercept does not disrupt the normal operation of the application. Optionally, the audio stream capture program takes the input and output audio streams and passes them through audio mixer and audio compression programs to yield a condensed recording of the original conversation.