Alabo MacPepple-Jaja, 2026
My work begins with a simple question: what happens when the voices that shaped you don't exist in any dataset?
My father, Diepiri, was part of the African diaspora in Britain. His voice, and the voices of his community, carry tonal and timbral qualities that have no representation in machine learning. They are not in Common Voice. They are not in LibriSpeech. When generative AI models learn what a “voice” sounds like, these voices are absent from the training data and therefore absent from the output.
PERI is my response. I curated a 60-hour vocal archive and trained a RAVE model to learn its acoustic identity. The output is not a replica. It is a generative collaborator that carries the tonal character of the source while producing something new. The model preserves without copying. It remembers without repeating.
This work is grounded in post-colonial musicology. The question is not only “can we synthesise this voice” but “who gets to build the model, who decides what it learns, and who controls the output.” When we talk about ethical AI, we usually mean bias in classification or fairness in recommendation. We rarely ask who is missing from the training data entirely, and what it means to build tools that make those absences permanent.
My methodology is designed to be reproducible. The pipeline (corpus curation, consent framework, model training, real-time inference) can be applied to any audio archive where cultural sensitivity and consent matter. This is not just an artistic practice. It is a framework for institutions, labels, and archives that hold recordings they want to activate with AI without erasing the people behind them.
Beyond PERI, I build tools that sit at the intersection of machine learning and sound: full-stack generation studios, real-time audio plugins that respond to natural language, and generative systems for electronic and experimental music. The common thread is agency. I want the people who make sound to control how AI interacts with it.
I am interested in commissions, residencies, and collaborations with festivals, galleries, and institutions working at the intersection of AI, sound, and cultural preservation.