The MPEG-4 Standard provides a complete frame-work for hybrid coding of natural and structured sound information that permits the description of complete spatial environments with a low amount of data. A new platform, based on a virtual DSP, has been developed to optimize the MPEG decoding and pre-processing interface to feed an array of loudspeakers rendering spatial audio by wave field synthesis. The virtual DSP is optimized for superscalar processors and has an instruction set adapted to MPEG-4 advanced audio features. An actual system has been implemented on a general-purpose processor and on a parallel multimedia DSP.