The SoDA (Sound Design Accelerator) project aims at providing a flexible software environment for soundscape generation. Based on semantic information, it provides both an annotation schema and an annotated library of sound files that operates in relation to a generative system that delivers the final audio content. SoDA provides the user with various forms of interaction: from incremental assisted exploration of semantic and audio content in real-time to completely automated off-line soundscape composition. In this paper we describe the semantic and audio components and the various interaction modes.