As mathematical ideas from artificial intelligence become more accessible to implement on personal computers and cloud based services, they enter the arsenal of tools for qualitatively innovative forms of human expression. In the creative field of music composition, the initial use of artificial intelligence for imitation will soon give way to new paradigms of companionship where the artificial intelligence systems inspire and assist in the human creative process. This new era of artificial intelligence-driven musical instruments is an opportunity for musicians at all levels of skill and technological prowess across cultures. Accessibility of these early tools however has so far been limited to programmers and very
narrow musical domains with minimal control over personalizing the outputs produced.
This thesis proposes Living, Singing AI : an accessible and bespoke artificial intelligence composition system that bridges the gap between non-programming musicians and modern generative algorithms. It provides a framework to iteratively generate and shape sonic material using one’s voice as an input with the following affordances – a simple non-programming interface, an egalitarian framework of multiple large-scale music synthesis models for brainstorming, and a focus on each individual’s sonic ideas and aesthetic preferences to create a personal, intelligent composition model. Using this tool, the composer is able to use their voice to generate a wide range of novel sonic construction material using open source music synthesis models with intentionality, and to shape a bespoke intelligent composition system. Through peer workshops and application scenarios in practice we show the potential of democratizing artificial intelligent music tools as a brainstorming system to provoke and direct serendipity – an essential ingredient of creating music.
This Living, Singing AI system is designed with the following primary objectives:
- A brainstorming tool for composers to interact and query different kinds of AI models with just
Centered around an individual’s ideas and aesthetic preferences, the Living, Singing AI
system provides an egalitarian structure to iteratively interact with a collection of modern
AI models simultaneously with just the voice. This allows our system to be used without the
need for programming or accelerated hardware while enabling the individual to construct
novel sonic material generated from a unique vocalized query. This makes it accessible to all
kinds of musicians without limitations as to cultural, musical and technological background.
- A personalized and iteratively proliferating composition object (Organism).
Furthermore, the system allows the individual to shape the generated material creating a
personal, intelligent and evolving model unique to them. The system allows a musician to
create a personalized collection of this AI generated material unique to their vocalized input
with a collection of tools to select, order and reflect on the generated material that
iteratively grows. This bespoke composition object or Organism can consequently be used
to generate intelligent mutable compositions in both an individual and a collaborative