Maria CHOIR is a human-AI interactive musical artwork, created to be experimented in person through the act of singing.
Grounded on the interaction of both the visitor’s and the synthetic reproduction of the artist’s voices, the installation offers an AI-tool that reacts in real time, sonically and visually, to whatever the visitor sings, like an experimental echo-chambered acapella karaoke duet.
Structurally, Maria CHOIR is an installative experience, where the visitor is invited to sing along with the deep voice of the singer Maria Arnal, a timber transfer model trained with data of her singing voice, with her characteristic ornaments and melismatic runs. In a safe dark space, a center-based transparent plinth lets the visitor see the inner architecture of the installation, where the computers, the sound interfaces and the cables are located and lighted to be seen. Four speakers surround the visitors to embrace them with all the sounds displayed. On a screen, an AI-animated avatar leads the visitor through the whole experience, while reacting and transforming with their voice in real time. The visitor can sing for hours, if desired, and the AI-device will react differently every time.
Joining both voices (visitor and AI-model) in a real-time duet, harmonizing and accompanying the voice of the visitor with whatever they sing, Maria CHOIR unfolds a generative range of effects, midi-sampled voices and textures that surround the visitor in a choir of both their voice and a multiplicity of Marias’.
FRONT VIEW
SIDE VIEW
Following the singing duet chorale experiment, Maria CHOIR has another invitation to the visitors: recording their voice for 20 seconds. They are asked to sing whatever they want and, in order to proceed with the recording, they have to leave their email. They are informed that their voices will be included for the training of a collective voice model that will be used both to create an experimental musical piece and for research. Their mail is kept safely in order to contact them through all the process.
On the edge of a sound installation, an interactive fun musical game and an experiment on consent and training collective voice models, this AI-artwork aims to create and reflect on speculative ways of singing and listening, co-creating songs and textures with a merger of synthetic and physical bodies.
Maria CHOIR is an artistic experimental approach to modeling voices, not interested in automated singing or deep fake voice impersonation, but in generative musical human-AI interaction. The AI-artwork is not focused on reproducing exactly what Maria Arnal’s physical voice can do, but it is concerned about the synthetic’s potential of expanding that and the social interaction of the whole AI-device.
Furthermore, Maria CHOIR is also the first of other AI-artworks that are being developed by the artist with a S+T+Arts grant, in partnership with BSC, reflecting AI-explainability, human-AI interaction and new narratives of what a voice can be in the 21st century.
As a complementary activation of the installation, Maria Arnal also presents a series of live performative conferences, sonic meditations where she combines her voice with her synthetic clones, a synthetic choir and invites the audience to sing too in a collective improvisation. These performances are designed specially to explain how models have been trained, offering new perspectives to the black boxes of AI models, and are an invitation to listen to several discarded materials through the research process.
At the moment (February 2024), more than 90,000 people have visited this installation, within the exhibition "AI: Artificial Intelligence" at the Centre de Cultura Contemporània de Barcelona (CCCB). Already more than 11000 voices have been recorded in the installation and more than 1000 people have attended the live performances.
Maria CHOIR has been directed by Maria Arnal, funded by Fundación Española de Ciencia y Tecnología (FECYT), programmed in the Barcelona Supercomputer Center (BSC), programmed and designed by the studio axolot.cat Iván Paz and Lina Bautista, all the visuals created by JP Bonino, curated by Lluís Nacenta and developed and exhibited by Centre de Cultura Contemporània de Barcelona (CCCB).
Concept, direction and music production: Maria Arnal
Funding and Support: Fundación Española de Ciencia y Tecnología (FECYT)
Technical Execution and Programming: Barcelona Supercomputer Center (BSC), axolot.cat (Iván Paz and Lina Bautista)
Interactive Design: axolot.cat (Iván Paz and Lina Bautista)
Visuals and Artistic Design: JP Bonino
Curator: Lluís Nacenta
Host and Exhibition Space: Centre de Cultura Contemporània de Barcelona (CCCB)