Journal of Popular Music Studies Volume 38, Number 2, pp. 34–58
Abstract:
In July 2021, the electronic musician and scholar Holly Herndon launched Holly+, a vocal deep fake modeled on her singing voice. Holly+ is powered by an artificial intelligence engine trained on record ings of Herndon’s voice. Musicians were welcome to use the transformed audio commercially and simply credit Herndon and Holly+. They also could release their work with the official imprimatur of a Decentralized Autonomous Organization (DAO), a Web 3.0 entity tied to the Ethereum blockchain. My paper situates the Holly+ project against the backdrop of intellectual property law in the US, both as it stood in the late 2010s and 2020s and in light of the legacy of legally sanctioned appropriation and extraction of work by Black performers. The DAO framework Holly+ uses is unlikely to gain traction, and Herndon and Mat Dryhurst’s provocative invocation of “identity play” suggests an uneasy relationship to the racialized legacy of mimicry and impersonation in the history of American popular music. But the project stands as an important attempt at creating an artist-driven, grassroots effort at monetizing, authenticating, and protecting the intellectual property of musicians in the era of machine learning.
Download