AI-generated band Velvet Sundown is a Spotify hit, but is the music any good?

The National

19 hours ago

In normal times, Velvet Sundown would be a good news story.
They released two full-length albums only weeks apart, amassing more than half a million monthly Spotify listeners, all while their tracks landed on popular mood-based playlists. At a time when few new rock bands are breaking through, their arrival stands out. There's only one complication – the band isn't real.
At least, not in the traditional sense. There are no verified photos of all four members, no live shows, no interviews and no clear production credits. Everything, from the album art to the band bio, points to Velvet Sundown being fully AI-generated.
But the point of this review isn't to play detective and spot the musical equivalent of the em dash. It's to ask, even if this music was made by machines, is it actually any good?
Floating on Echoes and Dust and Silence feel less like distinct records and more like two sides of the same coin. At its algorithmic heart, Velvet Sundown is more a stylistic experiment than a creative expression. They evoke the warm, washed-out tones of 1970s Laurel Canyon folk – a hazy Americana sound informed by soft guitars, genteel percussion and warm ambience. The references are convincing. But as a listening experience, it wears thin fast.
Take Dust on the Wind, currently the band's most-streamed track. It's laid-back, mellow and competently arranged. The bassline rolls along gently, the percussion shuffles lightly behind the guitars and the whole thing lands exactly where it should. While the song has a definite vibe, it's not enough if that's all there is.
Drift Beyond the Flame and The Wind Still Knows Our Name follow similar patterns, and after a while, that samey-ness starts to set in. And after 20 songs of this, the question stops being about whether they are real and more about why they don't make me feel anything?
Part of the answer lies in the vocals. The singer (credited as Gabe Farrow) – or rather the simulated voice – is programmed to sound like a restrained crooner, somewhere between a diet Chris Cornell and Jeff Buckley, but without the risk.
Every note falls exactly where it should, like Tetris blocks. Just when a vocal line is begging to be lifted or break slightly, it stops flat as if the air's been cut. You don't hear breath intake, strain or any of the human cracks that gives a performance its vulnerability. The voice never truly soars, and maybe, for now, it can't.
The music across both albums, all 26 songs in total, carries the same uniform restraint. The titles suggest emotional weight – End the Pain, Smoke and Silence and Drift Beyond the Flame – but the lyrics rarely move beyond generalities.
While criticising an album for vague writing can feel like low-hanging fruit, it's harder to ignore when the genres referenced are built on a tradition of evocative lyrics that are often direct, searing or emotionally grounded.
End the Pain promises catharsis but never builds towards anything. Smoke and Silence is filled with empty slogans (raise your voice, break the chain / Sing for peace, end the pain) and Dust on the Wind, with its soft tone and strongest melody, drifts through pastoral scenes without direction.
Even in folk or Americana, genres often known for their ambience and intimacy, there's usually a sense of movement, of intriguing emotional drift.
Think of Neil Young's 1970 album After the Gold Rush, a genre cornerstone whose songs sway between togetherness and dissonance. It features tracks such as Southern Man that bristle with urgency, and Don't Let It Bring You Down, which drifts between melancholy and resolve.
Or take Joni Mitchell's 1971 album Blue, where A Case of You feels fragile and raw, like it could unravel at any moment. These songs and albums sound intimate, but never inherently inert. With Velvet Sundown, everything sounds nice, but nothing surprises.
And for music made by a system designed to predict, maybe that's the only extent it can currently produce. This is what makes the band's creator or creators - they haven't been revealed - choice of genre strange.
You'd think AI's full-throttled invasion into popular music would begin on more familiar terrain such as electronic dance music or hip-hop – music built on software, loops and programmed rhythm. But instead, Velvet Sundown is making guitar-based music and those limits are clear.
Rock, folk, Americana – are genres that rely on, or even revel in, human traits – timing that's slightly off, choruses that perhaps run too long and vocals that crack. They're messy in nature. For all the cliches about four chords and a chorus, guitar music works because it's imperfect. AI can sketch the outline, but it can't inject the feeling or attitude that pushes a song somewhere unexpected.
Which brings us to the broader problem, not with Velvet Sundown, but with the ecosystem they're presently thriving in. Their success is less about the quality of AI replication and more about how streaming has reshaped what listeners value in music. Playlists used to be about exploration and discovery, and now they are seemingly about consistency.
Mood-based curation, such as the unofficial Spotify playlist Good Mornings – Happily Positive Music to the Start the Day featuring Velvet Sundown, has flattened the sonic landscape to the point where a fake song can sit comfortably between works by real, era-defining artists such as The Beatles and Billie Eilish. The result is a listening culture increasingly valuing indistinction. Music becomes background and texture, not narrative or expression.
The reported calls by artists and industry to flag or ban AI bands such as Velvet Sundown – who are, unsurprisingly, back with another album next week – are understandable. But that's not the only answer. We don't need fewer AI bands, we just need more human ones. Artists that can create music that is, perhaps, as focused as Velvet Sundown but with the kind of idiosyncratic touches and emotional expression that only humans can conjure. It's those qualities, more than anything, that have a chance of breaking the algorithm.