The one-day symposium AI and Digital Innovations for Voice and Vocal Music took place on 7 March 2025, organized by the Institute for Advanced Studies in the Humanities (IASH) at the University of Edinburgh, initiated by IASH fellow Dr. Alexandara Huang-Kokina. This symposium marked the second event of the Creative Digital Dynamics series, which aims to provide a forum for peer feedback and networking opportunities for researchers, performers, and practitioners in the field. Following the symposium, a concert was held at Reid Concert Hall, featuring AI vocal performances and an AI-based chamber opera, both of which were open to the public through ticket reservations. The event received support from several organizations, including the Creativity, AI, and the Human research cluster at the Edinburgh Futures Institute, the Royal Musical Association, the Edinburgh Centre for Data, Culture & Society, the Music and Letters Trust (Oxford Academic), and IASH. From the information presented at the symposium, the event attracted a diverse audience, including composers, data scientists, entrepreneurs, performers, music researchers, and arts administrators. Additionally, students from the University of Edinburgh and other graduate students attended, exploring career opportunities of AI and music. I participated in this symposium as a musicologist, educator, and leader of a technical team guiding conservatory students in integrating AI technology into their performances.
The first session of the symposium featured three expert talks covering various aspects of the field. Dr. Hedvig Jalhed from Lund University and Mälardalen University presented her recent studies on characterization, casting, and thematic elements in AI-based opera. Dr. Francesco Bentivegna, an artist and lecturer in Digital Theatre at the University of Bristol, shared his insights on the myths of storytelling in voice and AI from philosophical and literary perspectives. Dr. Robert Laidlow from the University of Oxford discussed his ideas and the language models he developed for an AI-based vocal performance in collaboration with Dr. Stephanie Lamprea, which will also be performed in the afternoon concert. He reflected on the compositional processes using language models as worldbuilding oracles.
After the coffee break, the keynote session commenced. Professor Ricardo Climent, Professor of Interactive Music and Director of the NOVARS Research Centre at the University of Manchester, presented on the topic of creativity among data. He showcased his innovative game engine projects, including Noh-Virtual for Beatboxer and Digital Twin (2020), which features a live beatboxer engaging in a vocal duel with their AI digital twin, drawing inspiration from Japanese Noh theatre and challenging the boundaries of human-machine interaction. The fourth expert talk was delivered by Dr. Emmanouil Benetos from Queen Mary University of London, who discussed machine learning paradigms for analyzing vocal music. These talks sparked discussions about the importance of data sharing for fostering AI innovation in music. Participants debated whether there should be a commonly recognized platform for artists to share and access data, or if training data should remain the property of the individuals who collected it.
I attended the parallel sessions on ‘Compositional Practices & Reflections’ and ‘Machine Learning & AI-Enhanced Vocal Performance’, while the other sessions focused on ‘Legal & Economic Dimensions’ and ‘Ethics & Aesthetics of AI Voice’. Dr. Jemily Rime, a professor from the Guildhall School of Music and Drama, explored how the perception of control can alter our relationship with new technologies and their applications to a singer’s voice, exemplified by a voice model she trained. Dr. Olivier Pasquet from the University of London examined the intersection of sound and text through the lens of micro-rhythms, expressions, and prosody, along a continuum of meaning and abstraction. Ananya Panwar shared her insights on embodied practice and artistic agency in music creation, employing an autoethnographic approach that included indigenous music, environmental crises, and feminist phenomenology.
Both Dr. Saurjya Sarkar from Queen Mary University of London and Michal Radošinský from the University of Ss. Cyril and Methodius in Trnava presented on AI voice separation, focusing on the complexities of this technique and comparing existing tools, respectively. In fact, discussions at the symposium and personal conversations with participants revealed a significant interest in AI voice separation among attendees. I presented two vocal performance videos that incorporated AI voices and video generation, exploring how the singer’s identity is represented through an assemblage of voices.
The symposium was followed by a concert featuring three vocal performances that incorporated AI technology, as well as the premiere of an immersive science-fiction opera. The concert was a full house, with the highlight being the chamber opera AI Kaidan: Yūrei (Ghost) of the AI Empire, produced by OperAI and directed by Dr. Alexandra Huang-Kokina. The piece featured a score for piano, violin, cello, and four singers, narrating a story set in a future world with AI agents. Composed by Atzi Muramatsu, the music blended contemporary styles with traditional Japanese elements, while the AI components were prominently featured in the libretto and image projections. A key aspect of the production was its interactive element, inviting the audience to influence the story’s outcome. Attendees voted between two divergent endings—one involving vengeance and the other refraining from it. While the performers demonstrated the audience’s choice of vengeance, they also presented the alternative ending after the main story concluded. This opera served as a pilot showcase for a larger immersive opera project, exciting audiences about the potential integration of evolving technologies in future operatic works.
The symposium provided a valuable platform for participants to share insights as technology continues to advance. Common questions focused on specific techniques and the development of training models, with some information about AI technology in the music circle being new to certain attendees. For instance, Dr. Jemily Rime shared her experiences from the AI Song Contest, an annual international competition. Dr. Hedvig Jalhed outlined recent repertoires of AI-based opera, while Michal Radošinský introduced several free AI tools for voice separation.
The AI and Voice Symposium facilitated exchanges that helped practitioners understand the current state of technology and successfully achieved its original aim of promoting diversity in vocal representations through AI. The insights and resources shared during this one-day event were rich and practical, offering valuable guidance for those interested in exploring the intersection of AI and vocal music. Extending the conference to two or three days in the future would further enhance interactions and collaborations among participants.
Misty Choi, The Hong Kong Academy for Performing Arts