This thesis explores the use of formal grammars for generative music, with a focus on
representing polyphony within a single instrument through an alternative representation in
which each instrument can contain multiple voices. Instead of encoding notes as compound
terminal symbols, the proposed model decomposes musical information into pitch, offset,
and duration, which are generated separately and combined into complete musical events.
This decomposition allows monophonic, homophonic, and polyphonic textures to be
represented uniformly, without requiring special terminal symbols or workaround encodings. Polyphonic structures emerge naturally through synchronization between individual
components, and therefore, the system supports both multiple simultaneous instruments
and multiple voices within each instrument.
The model is based on right-linear scattered-context grammars combined into a rulesynchronized multi-generative grammar system, implemented as a Python console application that reads JSONC grammars, derives musical structures, and exports results into
various graphical and audio formats.