Category · 5 projects
Music & audio
Songs, podcasts, sound design, voice and restoration work where AI served the ear.
Projects in this category
Rep. Jennifer Wexton's voice
Jennifer Wexton & ElevenLabs
A congresswoman who lost her speech to a neurological disease addressed the House of Representatives in a clone of her own voice, built from recordings made before her diagnosis.
Why it is not slop. Voice cloning as accessibility, used by the person whose voice it is, in the most public room in the country.
Elf.Tech
Grimes
Grimes released a tool that turns any recording into her voice and promised fifty percent of royalties to anyone who makes a successful song with it.
Why it is not slop. The opposite of the unauthorised AI Drake track that prompted it: consent, credit and a split, written down before anyone pressed generate.
Now and Then
The Beatles
The last Beatles song, finished from a 1970s John Lennon demo after machine learning separated his voice from the piano.
Why it is not slop. AI as a restoration tool, used only to recover something that already existed. The clearest case of technology serving the work.
Lost Tapes of the 27 Club
Over the Bridge
New songs in the style of Nirvana, Amy Winehouse, Jimi Hendrix and The Doors, generated to talk about mental health in the music industry.
Why it is not slop. The imitation is the argument: these artists should still be alive to write their own songs. The campaign is about the absence, not the output.
PROTO
Holly Herndon
An album recorded with Spawn, a neural network Herndon trained on her own voice and on a choir of collaborators.
Why it is not slop. The training data is consensual and the model is a collaborator with a credited role. The record sounds like nothing else.