Sound is an essential entity for creating realistic immersive virtual environments in applications such as computer games, health psychology, and scientific visualisation. Current multimedia libraries offer many tools for processing and synthesising sound but are usually executed on the CPU and/or suffer from vendor lock-in. In this paper we present a novel low-spec GPU implementation of an audio library. Preliminary studies suggest that the implementation improves speed by an order of magnitude compared to a CPU implementation. Furthermore a GPU-based implementation frees valuable resources on the CPU for other applications and it is ideal for applications where sounds and graphics are integrated and hence can be both executed on the GPU without requiring expensive data transfer between main memory and the graphics card.