GenCAST · Generative Compositional Assistant for Sketching Textures

GenCAST explores how artificial intelligence can support musical creativity by addressing a key challenge in human–AI interaction: the question of authorship and control. It focuses on musical texture — the way elements such as melodies, rhythms, timbres are layered and combined in time, from a single melodic line to complex polyphonic structures — as a central dimension of musical modelling. The project investigates how symbolic representations of texture can inform the design of a composition assistant that enables meaningful and controllable musical transformations.At a time when AI tools are reshaping creative workflows, recent advances — from generative audio models to real-time improvisation frameworks — have expanded the ways musicians interact with technology. Yet most recent industry solutions rely on opaque, prompt-based processes with little connection to symbolic musical structure. In response, GenCAST proposes a musician-centred assistant grounded in symbolic modelling, bridging AI and musicological knowledge.The project will curate a symbolic dataset, train a variational autoencoder (VAE) model, and integrate it into a composition plugin enabling composers to explore, combine, and reshape textures interactively. Validation with users will ensure both scientific and artistic relevance.The fellowship will be hosted at Maynooth University under the supervision of Prof. Victor Lazzarini, whose team combines musicological research with software development — an ideal context for mutual exchange with the fellow's expertise in texture modelling and AI. A secondment at STMS-IRCAM (France) under the supervision of Dr. Gérard Assayag will connect the project to a leading academic and artistic communities. This project will establish the fellow's profile as a researcher working at the intersection of musicology, technology and creative practice, enhancing their prospects for future roles in academia and industry.