Expanded Voices explores how AI voice synthesis reshapes identity, blurring the lines between self, body, and technology. The project focuses on the latent space of AI models—a multidimensional structure that encodes and compresses training data. In this digital reflection of a physical trait, identity can be shaped, replicated, and disembodied, revealing how Artificial Intelligence introduces new ways of understanding and expressing the self.
The installation invites the audience to speak or sing into a microphone and hear their voice transformed into that of artist Maria Arnal, while watching their vocal imprint appear in the latent space of an AI timbre transfer model—one of the technologies featured in Arnal’s latest show, AMA. This interactive experience, centered around the concept and structure of latent spaces, reveals how unsupervised algorithms model and manipulate voice. It offers a glimpse into the inner workings of AI, while prompting reflection on authorship, embodiment, and the fluid nature of identity in the digital age.
Evolving from a S+T+ARTS residency, the project is part of an ongoing collaboration between Maria Arnal and the Barcelona Supercomputing Center. It demonstrates the power of combining scientific and artistic approaches to drive innovation—opening new possibilities for healthcare, live performance, and the creative exploration of AI as both a tool and a subject of study.
