ABSTRACT This paper presents a new non-standard technique for waveform synthesis in the time domain using L-systems, a formalism related to grammars, fractals and automata. This technique, developed as part of a larger-scale compositional system, is based on waveform segmentation and offers various methods for generating wavetables. The paper first introduces L-systems and some specifics of their interpretation and discusses extensions such as incorporating genetic algorithms and designing hierarchical L-systems and L-system networks. The second half describes the implementation model in detail, proposes some sound synthesis strategies and presents paths for further work.