The use of non-intrusive virtual environments is gaining more and more importance but was focused mainly on addressing the visual sense. However, the human perception consists not only of visual input and thus it would be worthwhile to create multi-modal and interactive virtual environments. This thesis describes the techniques required to include the acoustic component into a virtual environment and furthermore the implementation of a software system, which takes advantage of these techniques to create complex acoustical scenes in real time. The system is based on the binaural technology. It features spatially distributed sound sources which are utilized to create an environment that is as authentic as possible. This comprises a description of the source, including its relevant angle-, distance- and time- dependent radiation, the sound distribution in the virtual scene (room acoustics), the perception-related consideration of all sound field components, as well as the exact reproduction of the artificial sound at the ears of the user. The focus of this thesis is put on the reproduction technology. In this context, an approach for dynamic crosstalk cancellation is presented, which enables a loudspeaker-based reproduction. The required filters are processed in real time on the basis of the position data and measured transfer functions of the outer ear. Furthermore the integration of this spatial audio system into a five-sided Virtual Reality display system is described and evaluated.