We’ve been acquainted with fake bands for years, animated proxies for real-life human meat bags who compose and perform the songs: The Archies, Prozzak, and Gorillaz are just a few examples.
If we want to get a little more complicated, there’s Japan’s Hatsune Miku, a holographic creation capable of performing live concerts using something called Vocoloid singing synthesis technologies developed by Yamaha. But she wouldn’t be able to perform without human-created music.
What do they all have in common? We know that they’re fake. Today… well, it’s getting difficult to determine what we’re being fed.
When we started talking about artificial intelligence a few years back, some openly wondered how long it would take before we had an AI music star. Technically, it seemed possible, if not for the fact that AI is not very intelligent. It’s a very powerful form of mimicry, but incapable of the emotion required to write music convincing to the human soul. Any music an AI program generates is based on analyzing millions of datapoints fed to it in the form of music written by flesh-and-blood musicians.
I read somewhere — and I’m paraphrasing — “Until an AI program falls in love and goes through a painful breakup, it will be incapable of writing an original song like Tammy Wynette’s D-I-V-O-R-C-E.”
There have been attempts, of course. In 2019, there was a brouhaha over FN Meka, a virtual rapper who amassed over 10 million followers on TikTok and picked up deals and partnerships with Xbox and Amazon. FN Meka was even “signed” to Universal Music before people began attacking the creation for its appropriation of Black culture, the use of stereotypes, and how this thing like to drop the N-word. Universal dropped FN Meka and had to issue an apology. But this thing wasn’t autonomous. FN Meka was built through a collaboration between artists, singers, producers, coders, and programmers.
But that was then. Now we’re living in an era where generative AI is much better at faking things in a convincing manner. Programs like Mubert, AIVA, Soundful, Anthropic, and a dozen others can not only spit out full songs based on simple text prompts but also create a music video to go along with it. And it takes just seconds.
Musicians are rightfully concerned about this. Their labour is being used to train their robot replacements without their permission and any compensation. But still, it’s just AI, right? And we can tell — feel — when we’re presented with an AI song. We humans will never accept full AI artists.
Get breaking National news
For news impacting Canada and around the world, sign up for breaking news alerts delivered directly to you when they happen.
Well…
There’s been much talk over the last month about Velvet Sundown, a rock band with four grungy-looking members. The group racked up close to 750,000 monthly listeners on Spotify in less than a month, an impressive accomplishment for any act, let alone a brand-new one. They’ve already released two albums with a third set to drop on July 14. That’ll be a total of 39 songs in five weeks. Speculation is that this is a fake creation.
Someone (or something) has posted on X blasting critics for assuming that Velvet Sundown was AI. In a taunt to music journalists, the post says, “Unbelievable the number of journalists who simply don’t check in with the **real people** their coverage affects. Guessing most of these ‘journalists’ are actually just AI!” Another post reads, “They said we’re not real. Maybe you aren’t either.” And another: “Many news outlets are falsely reporting that we are an AI-generated band,” another message states, “and nothing could be further from the truth! #VelvetSundown #NeverAI”
Yet anyone who has attempted to reach the band has been unsuccessful. The domain velvetsundown.com leads to a very dodgy-looking squatting site promoting plastic surgery.
Deezer, the French streaming site, has slapped an AI label on Velvet Sundown, saying their new AI-detection tech has found “some tracks on this album may have been created using artificial intelligence.” Whoever (whatever?) is behind the group was miffed. “We have reached out to Deezer to get them to remove the AI-generated label which they wrongly applied to our music. They said they will do it.” So far, though, nothing.
Music Ally, a music industry site that’s been investigating the situation, asked French AI detection company Ircam Amplity to run Velvet Sundown tracks through its BS detector. When the analysis came back, 10 of the 13 songs were declared AI-generated with a confidence score of 100 out of 100. Two scored 98. One hit a semi-convincing 73. Analysis of the artwork and band photos of the group point to them being AI-generated. For example, one group picture features anomalies like an incorrectly stringed guitar and a Stratocaster with no apparently pickups.
What’s powering this? The best guess from Ircam Amplify is V4.5 of a model called Suno. This points to someone who has mastered the finer points of Suno in a very impressive way. And with three-quarters of a million listeners on Spotify, this could turn out to be rather profitable. But it could also be just a proof of concept for the powers of AI and songwriting, an artsy prank, or just someone with too much time on their hands. Some have even suggested that this could be a clever marketing scheme launched by a real band.
The truth? It’s a hoax. A band “spokesperson” has said as much. An “art hoax,” they said.
All this would have been more impressive if Velvet Sundown had music that was in line with today’s musical zeitgeist. Instead, it’s passable, inoffensive stuff that wouldn’t have sounded out of place in the ’70s or ’80s. It’s unoriginal and, if I have to say it, bordering on boring. (Try Interlude on Spotify.) Then again, so is a lot of today’s contemporary pop.
It would be one thing if Velvet Sundown was an outlier. It is not. Check out Aventhis, an outlaw country artist with no apparent corporeal form. Yet he — it — has over a million Spotify listeners each month to his/its 57 tracks with one song racking up two million plays. (There is some kind of real-life lyricist behind this named David Viera, but he says, “(The) voice and image is created with the help of AI. The lyrics are written by me.”
Then there’s The Devil Inside, another outlaw country…thing with 700,000 Spotify listeners. A track entitled Bones in the River has nearly two million streams. Everything about The Devil Inside is clearly fake, including the images on the official Instagram page. Want merch? It’s available, too. And once you start listening to entities like these on Spotify, don’t be surprised if the algorithm throws a few more your way: Nick Hustles, The Smoothies, Velvet Funk, King Willonius, Hyperdrive Sound.
So what do we make of this? Is this more AI slop? Will the streaming music platforms step in further? What sort of copyright infringements might be involved? Is this the tip of the iceberg when it comes to music fans embracing music made by robots who have stolen the souls of real-life musicians? Perhaps, though, others will see this as just another music-making tool.
We’ve reached the point with AI where we can no longer trust our eyes or ears. What’s real anymore?
© 2025 Global News, a division of Corus Entertainment Inc.
