Bla Blavatar vs Jaap Blonk

An experimental collaboration between Jonathan Reus and the virtuoso vocal improvisor Jaap Blonk, exploring voice dataset making as a laborious, improvisational and absurdly banal performance form.

The performance is an absurdist take on the cultural obsession with generative AI avatars and the goldrush to capitalise on creative automation. Rather than focusing on the uncanny realism of voice clones, the performance foregrounds the physical, creative and mental effort of performing sound poetry precisely according to an AI-friendly score, inspired by traditions of phonetically balanced reading scripts in speech research, and artistic traditions of typographical musical scores in sound poetry.

Reus performs as BLA BLAVATAR, using a custom real-time voice synthesis instrument called Tungnaa, which is trained on previous dataset poem performances by Blonk. The dynamic between Blonk and Reus sways from moments of precise and controlled material to improvised duets between Blonk and his vocal avatar. During the performance all of Blonk’s vocalisations are recorded and added to a growing public dataset, and will be used as training data for making BLA BLAVATAR a better public speaker. Together, Blonk and Reus bring the vocal labor behind AI voice models to the stage, as well as use datasets and dataset making as a vocal art form that is fundamentally creative and exploratory, rather than a dehumanising means to an end.

Credits:
Bla Blavatar vs Jaap Blonk is one outcome of the 2023-24 S+T+ARTS AIR residency D̴̼̥̉a̸͂͜ͅd̵͂͑ͅą̶̀͐s̵̯̈́e̶̜͔̍͝t̶̘͎̃s̶̫͙͂, by Jonathan Reus. This performance was made possible with support from:

PiNA – Association for Culture and Education
Stroom Den Haag
S+T+ARTS AIR  DaDaSets

Footage: Dan Ortinez Marti
Editing: Flora Reznik

Tungnaa was developed by Jonathan Reus and Victor Shepardson with additional R&D support from the Intelligent Instruments Lab, University of Iceland