Velvet Sundown

So a few months ago, a new band released some music on Spotify (and elsewhere) and gave people something new to get upset about. The band, Velvet Sundown, "is a synthetic music project guided by human creative direction, and composed, voiced, and visualized with the support of artificial intelligence." In other words, the band does not exist, and the music is AI-generated. The music is not horrible, it's insipid 70s-inspired folk rock and is somewhat pleasing to the ear. In June 2025 they rapidly gained over a million streams on Spotify, and four months later they still have 250,000 monthly listeners. 

And the world exploded. First it was "wow a mysterious new band, but they're kind of weird" but it quickly turned into a consensus that this was the beginning of the apocalypse. Everyone from the New York Post to Rolling Stone to the Washington Post to the BBC to the Guardian to NPR weighed in. Reddit and X blew up. It attracted a hoaxer who falsely claimed to be responsible for the band, just to attract attention himself. Rick Beato, a music analyst with over 5M followers on YouTube and another 1.4M on Instagram, ominously started his story with "So it begins..." and raised legitimate concerns about authenticity and human connection, streaming and discovery mechanics, and the risk to the ecosystem. 

The main worries seem to be that music could increasingly be produced by “machines trained on hits” rather than humans crafting songs, which may flatten diversity and innovation. And the AI models are likely trained on human-made music without attribution or payment. Moreover, you can produce a hit-like song in 15 minutes, so you can completely flood the music platforms and crowd out "real" music and suck up what little royalties there still are away from human creators.

Was the moderate success of Velvet Sundown a tribute to people's curiosity about this horrible degeneration (the next Vanilla Ice) or a genuine and worrisome inability to differentiate "real" hit music from computer-generated drivel? Was it a Turing-test moment in music and do people (other than those writing or vlogging about music) just not give a shit who or what made their music, as long as it has a pleasing sound that matches our formulaic and predictable listening behavior? Will this mean that increasingly, future hit music will be fully derivative?

From the beginning of recorded pop, producers have manufactured acts, assembling talent, image, and story to fit a commercial goal. Think The Monkees, The Partridge Family, Menudo, New Kids on the Block, the Spice Girls, NSYNC, K-pop. German producer Frank Farian had enormous hits in the 70s and 80s with Boney M and Milli Vanilli, in which none of the "band members" had anything to do with the music (written, produced, even partially sung by Farian), raising questions of deception and authorship. Vanilla Ice was marketed as a tough, streetwise rapper from Miami when that persona was constructed by his label; authenticity backlash was huge once the marketing fabrication came to light. The Gorillaz were authentic in their music but a virtual band consisting of animated characters. AI-generated bands like The Velvet Sundown are really an extreme continuation of that tradition, although it removes the last remaining layer, the human performers themselves.

One part of the core debate is unchanged: authenticity. What makes music "real"? Who really made this? Does authenticity matter if you like the music? AI just pushes this old debate into existential territory: What if there is no artist at all? Or is the person directing the prompts now the artist? And how is that different from Frank Farian, with just a better synthesizer?

Every major technological leap in music (the electric guitar, multitrack recording, drum machines, synthesizers, samplers, Auto-Tune) initially drew criticism as fake or soulless. Technology has steadily taken on more of the creative load in music production. The difference is one of degree and autonomy, an evolutionary extension. But did we cross some major threshold, like creating new musical ideas autonomously, not just processing existing sounds, the Judgment Day of music? Did Skynet just become a self-aware artist? 

I am not too upset by Velvet Sundown. YouTube has already made a new policy to de-monetize AI-generated content, taking the financial incentive out of flooding the platform with auto-generated crap and saving money while “protecting creatives.” I’m guessing Spotify will, too; who will argue? Perhaps Spotify introduces a new filter that lets you exclude generated music. Maybe there will be labeling or acknowledgment requirements. Or maybe it will make music better?

And this is not the first time a disruption was rumored to be the end:

• When Edison's phonograph appeared, "no one will see live music again"
• When radio emerged in the 1920s, "record sales will collapse"
• The electric guitar was "the end of real musicianship"
• Synthesizers: "Robots have replaced musicians"
• The Sony Walkman: "Music will become antisocial"
• Sampling and hip-hop: "Stealing, not creating"
• Napster and easily copied mp3s: "The death of the music industry"
• The iPod and iTunes: The death of the album and the record industry"
• Spotify and streaming: "Artists will starve"
• Auto-tune and digital production: "The end of real singing"
• And now AI and virtual artists: "The death of human creativity"

We are deep in the era of algorithmic optimization. Since the beginning of the streaming era, producers have been writing songs to fit the preferences of the algorithm: short intros, shorter durations, mood playlists, a powerful hook, ... Machine learning already curates, recommends, and sometimes co-writes through AI mastering and auto-arranging and recommendation tools. Producers study data like first 7-second retention, playlist inclusion probability, and danceability scores. Songs are built for micro-moments and viral loops, not for albums or narrative arcs. AI-assisted, data-tested music is the new arena rock: huge, efficient, and a little hollow.

It feels like we are in a deep cultural loop that repeats every few decades. What makes music "good"? What makes new generations fatigue from the uniformity of mainstream music?  Historically, when music becomes too homogeneous and polished, data-driven, and risk averse, audiences rebel and cultural resets happen. A new wave rejects polish and re-injects vulnerability and imperfection. Late 1950s teen pop was disrupted by the 1960s rock counterculture. The Hit Factory era of the 1970s (slick, predictable, radio-friendly hits that maximized airplay and record sales) degenerated into disco and led to the backlash of punk and new wave. 1980s glam and hair bands and MTV pop and arena rock created the backlash of grunge and rap in the 1990s. Boy bands, alt rock, EDM and hip hop, a sea of pop homogeny and the indie search for authenticity. The specifics change but the emotional physics don't. 

So maybe this mediocre AI band is not the end of the world. Maybe it's just time for another backlash. 

Would be cool to see the next Nirvana.

(Oh, by the way, this post was written largely by ChatGPT).