Holly+

An Artist Becomes an Instrument

In 2021, the experimental musician and artist Holly Herndon launched Holly+, a project that fundamentally reframed the relationship between an artist, their creative identity, and artificial intelligence. Holly+ is a deep learning model trained on Herndon’s voice — her singing, her speaking, her entire vocal identity — that allows anyone to upload audio and have it transformed into something that sounds like Herndon singing.

The concept is disarmingly simple. The implications are profound.

Any person, anywhere in the world, can take a piece of audio — their own singing, an instrument recording, even ambient sound — and run it through the Holly+ model. The output is that audio reshaped into Herndon’s vocal timbre, phrasing, and character. The original melody, rhythm, and emotional contour remain, but the voice becomes Holly’s. The result is music that Herndon never performed, never heard, and never approved — yet is unmistakably her.

Why Herndon Did It

Herndon’s decision to create Holly+ was not reactive. It emerged from years of engagement with the intersection of technology, voice, and identity. Her 2019 album PROTO, created in collaboration with an AI system she called Spawn, had already explored how machine learning could participate in musical creation. Holly+ was the logical next step: not just collaborating with AI, but offering her own creative identity as a medium for others to work with.

In Herndon’s framing, Holly+ is an experiment in what she calls “data dignity” — the principle that people should have agency over how their data (including their voice, likeness, and creative identity) is used in AI systems. By voluntarily creating and releasing a model of her own voice, Herndon established a framework based on consent and intentionality, in sharp contrast to the AI models trained on scraped data without artist knowledge or permission.

“I wanted to create a positive example,” Herndon has said, “of what it looks like when an artist engages with AI on their own terms, rather than having AI imposed on them.”

The Governance Model

What makes Holly+ genuinely innovative — beyond the technology itself — is its governance structure. Herndon established a DAO (decentralized autonomous organization) to manage the use of her voice model. The DAO includes a committee of artists, technologists, and community members who review submissions and decide which Holly+ creations are officially endorsed.

This governance layer addresses one of the central anxieties of the AI age: if your creative identity can be replicated, who controls that replication? Herndon’s answer is a participatory structure where the artist retains meaningful agency without exercising absolute control. It is a prototype for a new kind of creative rights management — one designed for a world where identity itself can be digitized and shared.

Revenue generated from Holly+ creations is distributed through the DAO, creating an economic model where Herndon benefits from the use of her voice even when she is not the one using it. This is a direct response to the reality that most AI voice and image models generate value for platform owners while the artists whose data powers them receive nothing.

The Art and the Music

The Holly+ outputs range from haunting to humorous, from musically sophisticated to deliberately crude. Some creators have produced genuinely beautiful compositions that use Herndon’s voice in ways she might never have imagined. Others have created absurdist experiments — feeding bird songs, power tools, or political speeches through the model to hear what they sound like in Holly’s voice.

This range is the point. By relinquishing control over how her voice is used — within the governance framework she established — Herndon created a living, evolving artwork that grows with every contribution. Holly+ is not a fixed piece. It is a creative ecosystem, a platform for collective experimentation, and a provocation about the nature of artistic identity.

The project has been exhibited at major cultural institutions including the Serpentine Gallery in London and featured at Ars Electronica. It has been discussed in academic contexts ranging from music theory to intellectual property law to AI ethics. Its influence extends far beyond the music world because it addresses questions that every creative discipline will eventually face.

The Deeper Questions

Holly+ forces a confrontation with questions that most of the AI art discourse avoids:

What is a voice? Is it a physical attribute, like the shape of your face? A creative instrument, like a learned skill? A component of identity, like your name? Holly+ suggests it is all of these — and that each framing implies different rights and responsibilities.

What does consent look like in the age of AI? Herndon consented to the creation and distribution of her voice model. Most artists whose work trains AI systems did not. Holly+ makes the difference vivid and demonstrates that consent-based AI art is not just ethically superior — it is also more artistically interesting, because the artist’s intentional participation adds layers of meaning that scraped data cannot provide.

Can an artist exist as a platform? Herndon has described Holly+ as a way of existing in multiple places and creative contexts simultaneously. She is no longer just a performer who makes music — she is a creative infrastructure that enables others to make music through her. This is a radically new model of what it means to be an artist, and it may prove more durable and more interesting than the traditional model of the solitary creator.

Holly+ does not resolve the tensions around AI and art. It inhabits them — deliberately, thoughtfully, and with a generosity of spirit that much of the AI discourse lacks.