Who does your voice belong to? For musician Holly Herndon, the answer is ‘everyone’

For Document’s Winter 2021/Resort 2022 issue, the musician envisions the future of intellectual property in the era of vocal deepfakes

I’m on Zoom with experimental musician Holly Herndon, and her deepfake twin is singing to me. This is Holly+: an AI-driven instrument that takes audio recordings and reinterprets them in her voice, making disembodied collaborations like the one I’m hearing—a hauntingly realistic cover of Dolly Parton’s “Jolene,” sung by Herndon’s vocal model—not only possible, but available for others to experiment with via her website.

When we speak, Herndon is slated to debut the new version of Holly+ we’re listening to, which has the capability to transform audio in real time. “My partner Mat [Dryhurst] can sing through my voice. When we first got it working and he was doing it, I had to leave the room because I was like, This is too creepy. It was my voice, but with a hint of his British accent,” she laughs. “At the same time, it’s so exciting to imagine all the different voices that one could perform through. With new technologies, all these questions open up: What is vocal sovereignty? Who owns the voice?”

While many musicians have shied away from artificial intelligence, Herndon was an early adopter. Her 2019 album Proto was the first to utilize singing neural networks, and saw Herndon collaborate with her “AI baby,” Spawn, to create atmospheric soundscapes populated with an ensemble of human and inhuman vocals. At once intensely cerebral and emotionally intimate, the album’s sonic underpinnings trace back to her earlier work, which utilizes a palette of avant-garde sampling techniques to unite the synthetic and sensual—not only demonstrating the potential of emerging technologies, but often humanizing them in the process.

By releasing her own voice to the public, Herndon aims to highlight the challenges these technologies pose while also setting her sights on collective solutions that could safeguard creators against exploitation. “Vocal deepfakes are here to stay,” she said in a statement accompanying Holly+. “A balance needs to be found between protecting artists, and encouraging people to experiment with a new and exciting technology.”

“Technology will change things, but I refuse to become cynical, and I refuse to give up my agency in this process. I’m not scared; I’m excited about what this new technology could unlock.”

Dress by Ottolinger. Boots by Chloé.

One common issue, Herndon explains, is that the infrastructure to protect creators often comes after a new medium is popularized. She cites the advent of recording technology as one such example. “It opened up entirely new genres of music [through sampling], but it also had its pitfalls, and creators were screwed over by it,” she says, referencing the famous “Amen Break”—a now ubiquitous six-second clip whose original creator, the drummer Gregory Coleman, died homeless, never having received royalties for the most-sampled piece of music in modern history.

Herndon addressed the need to develop new ideological, political, and technological structures in her second studio album, Platform (2015), sampling the domestic and digital sounds of her daily life to create galvanizing experimental tracks that act as a rallying cry for new fantasies and forms of strategic collective action. In the months since NFTs entered the mainstream, blockchain has become a home for much of this new community infrastructure. “There are a lot of on-ramps being built in that space right now,” Herndon states, explaining that the protocol originally popularized by cryptocurrency is now being used by creatives to keep and track digital certificates of their intellectual property. Holly+ is currently overseen by a decentralized autonomous organization, or DAO: a democratic collective that licenses out the official use of Herndon’s voice to approved artists, meaning that each creation can be traced back to its original source. Herndon and Dryhurst have also launched a digital auction house where artists can upload and sell works made with her likeness. The proceeds are then funneled toward the development of new tools, which artists can use to create and sell more works—it’s what Herndon terms a “virtuous cycle,” where individual creativity serves to benefit the well-being of the collective.

The idea of creative symbiosis is one she explores on Interdependence, a podcast Herndon hosts with Dryhurst. By engaging with preeminent voices across art and technology, the pair effectively map out the challenges of our current cultural upheaval—a moment characterized by the ever-expanding technosphere and a looming conflict between what graphic designer David Rudnick, a recent podcast guest, terms “virtual prime” and “physical prime.” Rudnick describes this impending paradigm shift as a struggle between two value systems: the set of beliefs held by a generation of individuals who experience their sense of purpose, self, and future as largely tied to IRL presence, versus the principles favored by an emerging network of digital natives who identify online interactions as a key source of value and meaning in their lives (and thus might choose to invest in cryptocurrency and NFTs over physical property and art objects).

“I’ve been experimenting on my own voice and image, because it seems like the most ethical thing to do…As we move into this new reality of machine learning, ideas around intellectual property are going to shift.”

Herndon herself has long been aware of NFTs, but has only recently started producing and releasing visual art. In her debut NFT collection, Crossing the Interface (2021), Herndon and Dryhurst utilized machine learning to generate nuanced visual compositions incorporating text from Iranian philosopher Reza Negarestani. “Crossing the Interface was originally a libretto that I commissioned from Reza [years ago], because I had this idea of doing a decentralized live performance,” she says. Held at the Guggenheim in New York, this disembodied performance featured a remote performer whose voice was made to travel through space using a configuration of speakers called an ambisonic array. “Physically, she was in Oklahoma—but it was almost like she was hyper-present, because the sounds of her body and footsteps stomping through the audience were so magnified. I was trying to play with this idea of technologically-mediated situations being very intimate, by having her feel even more present than if she had actually been there,” Herndon recalls, noting that this original performance took place five years ago: a time when many still rejected the idea that digital interactions could possess the same emotional weight as physical ones.

Herndon was inspired to revisit this idea after witnessing the rapid development of the NFT art space, following the record-breaking sale of Christie’s first fully digital artwork. “We released Crossing the Interface as an NFT because it touches on the same ideas that Web3 is talking about: decentralization, the importance and value of digital work, and our digital-emotional selves. So we went back and took the libretto that we commissioned from Reza years ago, and we broke that into small pieces, then used part of the text along with our own aesthetic prompts to create the animations,” she says, explaining that machine-generated visuals often entail an element of experimentation on the part of human creatives.

Shirt by DIOR RE22. Pants and skirt by Aileen Klein.

In her recently released series of self-portraits, CLASSIFIED, Herndon takes this collaboration one step further, challenging a neural network called CLIP to assemble a visual representation of who she is using a series of custom prompts. “When we started, CLIP already had an idea of who I am,” Herndon says. This is because her existence had already been documented in public data sets—the collections of images, sounds, or text that are used to train machine learning algorithms. “It’s hard to exist online without being swept up into a training set,” she explains. “I’ve been experimenting on my own voice and image, because it seems like the most ethical thing to do. But there is currently no intellectual property law around training sets—and as we move into this new reality of machine learning, ideas around intellectual property are going to shift.”

Even with this in mind, Herndon’s perspective is surprisingly optimistic. “I think it’s important that we approach [new technology] with caution and with criticism, but not with cynicism,” she says, explaining that while it’s tempting to criticize new developments for their potential misuse, halting technological advancement is a losing battle. Instead, she wants to empower individuals to experiment with new technologies by improving the social and political infrastructure that surrounds them. “Technology will change things, but I refuse to become cynical, and I refuse to give up my agency in this process,” she says. “I’m not scared; I’m excited about what this new technology could unlock. But I also care a lot about creator compensation and the health of our communities, so I want us to take some lessons from the past and hopefully implement better systems moving forward,” she says. “At the end of the day, it’s much easier to just criticize something than it is to propose something.”

It’s a sentiment that echoes the observation of feminist scholar bell hooks. “When we only name the problem, when we state complaint without a constructive focus or resolution, we take hope away. In this way critique can become merely an expression of profound cynicism, which then works to sustain dominator culture,” she wrote in 2003, stating that both recognizing the problem and articulating what we can do to address it are necessary to inspire change. Taken in context, optimism is not a sign of naivete, but a form of resistance—and it’s this hope for a better future that animates Herndon’s work. “People often forget that machine learning is not alien intelligence; it’s aggregated human intelligence,” she says, speaking of the capacity of emerging technology to hold a mirror up to culture: reflecting the challenges we face, and creating new opportunities for collective solutions. “Much like the voice, artificial intelligence is inherently communal—and we’ve all contributed to it through participating in humanity.”