PercAI: A Generative AI Model for the West African Balafon Polyrhythm
PercAI is an explorative project aimed at developing a human-AI co-creative model to enhance the creativity, analysis, and learning of West African polyrhythms, specifically the traditional balafon music composed by Youssouf Keita and Aly Keita. Through the use of AI technology, the project seeks to inspire both African and Western musicians, as well as bringing new insights to ethnomusicology, and the global music creatives on the use of AI. As continuation to Creatie Research Group’s AI Musicking project (2023-24), PercAI will research into the ‘bottom-up’ approach (machine learning ML and deep learning) to learn from data and discover the underlying patterns among the balafon polyrhythms, capturing the compositional relationships among melody, patterns, and pulse. The process will train a generative AI model—a human-AI co-creative model prototype—that is capable of producing rhythmic and melodic patterns to accompany a user-inputted main melody, hence, empowering performance and creativity in balafon style. Situated at the center of the research, the balafon musicians will be invited to record and sample the balafon compositions in MIDI format (i.e., data collection), feedback, and guidance to refine the AI model. Hence, the project embraces a participatory approach: feedback loops and co-creation concert comprising KCA students, traditional musicians and percussionists. Additionally, PercAI will investigate the ethics inherent in developing AI models for traditional music forms.
Results
• Accepted publication for the “Editors Forum: Ethnomusicology and AI” of the British Forum for
Ethnomusicology with title “The potential and ethics of AI in traditional music” (forthcoming 2025).
Through interviews with African musicians, I will reflect on the ethical issues of building and using AI
for traditional music
• Generative AI model prototype for analysis, creativity and learning of West African balafon polyrhythm The ultimate goal is to develop an algorithm that when musicians record their melodies with a simple recording device (i.e., smartphone), then, the AI will generate accompanying patterns and rhythmic
pulse in the style of balafon polyrhythm. It is a explorative project to investigate the feasibility of developing and using such AI model for the West African balafon polyrhythm and musicians. The
prototype AI model will explore the underlying relationship between the melody, patterns and pulse of balafon polyrhythm
• Short articles and public presentations submissions on various topics during this one-year project and
the future follow-up projects, including artistic research, AI technology, and ethnomusicology34
• Balafon workshop for marimba students at KCA
• The human-AI co-creation sessions and concert. It will be a collaboration of the balafon musicians,
percussionists, and the KCA marimba students
• Advices on AI ethics and copyright. Apart from musicians, I may consult the lawyer linked to the
musicology department of RMCA, SABAM, organisations Creatives Garage in Kenya and Morzilla Foundation, and an AI company Josplay based in East Africa.
