The author describes an Internet-based audio composition and diffusion system, Eavesdropping (2007–2008), designed for public spaces where several computer users are gathered, such as cafés. Compositions are created from abstract mood objects rather than musical structures. A composer uploads a set of audio files to represent the different moods in the composition. During a performance, a server-based Conductor selects audio files from this set to be played at each participant's laptop based on the composition, the number of participants in the room and the time they joined the performance. This project aims to enhance awareness of and connectedness among individual members of an audience at a generative musical performance by encouraging shared experiences.