How to Make a Viral AI Hit
Two months ago, an anonymous songwriter wearing a sheet over their head set the music world aflame when they shared a song they recorded on their own featuring an uncanny copy of Drake’s voice, made possible with artificial intelligence. “Heart on My Sleeve” — still imperfect, but catchy and close enough to convince many listeners it could be a potential Drake hit — caused a national stir as the music business and the rest of the world tried to figure out what to make of it. The song lasted just two days before Universal Music Group, the world’s largest music company, spoke out about infringement and the major streaming services removed it. But the point was made clear: AI music is here, and it’s not going away any time soon.
What has remained unclear, however, is exactly how these songs are getting made in the first place. The uninitiated might see these tracks as mind-blowing magic tricks where, at the push of a button, anyone can churn out a ready-made hit by a name-brand superstar. But under the hood, it’s a much more complicated process that still requires a human being with songwriting and producing skills to be even halfway convincing.
To find out how these tracks are getting cut, and to test out the full potential of all of AI’s music tools, Rolling Stone asked Grammy-winning songwriter Evan Bogart to walk us through. Bogart — whose writing credits include Beyoncé’s hit “Halo” along with tracks for the likes of Demi Lovato, Britney Spears, Jason Derulo, and Jennifer Lopez — has been experimenting with beatmakers, voice clones, and other AI songwriting tools for months. The result of this experiment? In just a few hours, he was able to rustle up a fairly convincing Biggie rap about AI taking over the world featuring vocals from Harry Styles for the hook.
Bogart has used ChatGPT to help him write lyrics, used AI beatmakers to kick off sessions, and tried out voice clones of artists including Rihanna to finish old demos he made where the artists he pitched didn’t think a song would work for them. He says it’s an effective option: “Sometimes people say, ‘Oh, I don’t think I can hear my artist on that,’ but now it’s like, ‘Let me show you.’”
Bogart rejects the notion that AI will ever fully replace human artists, noting that he sees AI as a tool to bounce ideas off of and get new perspectives, more like a co-writer. As the CEO and founder of independent music company Seeker Music — which has bought music rights from artists like Run the Jewels and Christopher Cross — Bogart also sees AI as a way to boost revenues from his catalog because it could make creating loops or samples much easier.
“I love the idea of collaborating with AI. There are certain things I gravitate toward naturally as a music creator, and maybe there’s other ways of writing I don’t know about that collaborating with AI might open me up to,” he says. “I think the people who embrace it in that way will find it really positive.”
In a two-hour writing session, starting from scratch, Bogart and co-writer Austin Max took Rolling Stone through the entire process. In this case, the prompt they set out for themselves was to make a song as close to the Notorious B.I.G.’s work as they could get. Feeling that the real Biggie likely wouldn’t have cut a particularly cheery song about the topic of AI, Bogart went dystopian, referencing the Terminator films and saying AI was spreading around the world like “carcinoma.”
Using an AI stem separator, Bogart and Max isolated some of the drums from DMX’s “X Gon’ Give It To Ya” (which Seeker has a stake in the publishing on), fast-forwarding what would be a much more time-consuming process if a producer didn’t have access to the stems themselves. They coupled the drums with a new AI-selected beat they made on music production platform Splice to make the base of the song’s backing track.
With the beat finished, It took Bogart just 10 minutes to write lyrics using ChatGPT, which helped him come up with rhymes and Biggie-style lines. He recorded his entire verse before plugging the raw vocals into the voice cloning software Uber Duck to replicate Biggie’s distinctive tone. They used another AI tool to clean up the recording and remove some of the distortions on the first, gravelly pass. The result still had some noticeable imperfections, but they were much harder to hear when they played the verse over the beat.
Looking for a hook, Bogart and Max stumbled through a few lower-quality clones for Ariana Grande, Rihanna, and David Bowie before settling on a still not-very-convincing but more passable Harry Styles, fused with Max’s voice this time to sing the phrase “taking over the world” on a loop.
It’s an impressive feat, if still a bit concerning to anyone who thinks this technology shouldn’t exist at all. But Bogart says the new frontiers that come with AI have given him more inspiration.
“I have so much fun when I make music — this brings the fun back into it, where I can explore and do something different,” he says. “People ask if I write on my own. I hate writing on my own. What I love about co-writing is I never know what a co-writer is going to come up with. What’s cool about AI is that it puts me back in a seat where I could write on my own again. AI is going to generate things I wouldn’t have thought of, and I can say things back.”
Bogart acknowledges that there need to be rules in place to ensure the music is used properly, and that the original creators are fairly compensated when AI songs are used commercially. He thinks AI may easily replace music written for purely commercial purposes, like ads or white noise — but even after making a brand new song by a long-dead rap legend in the amount of time it’d take to cook a gourmet dinner, he doesn’t see the tech driving out actual artistic creation. At least not yet.
“If they’re making production music for cues, for backgrounds of reality shows, I don’t know why you couldn’t use [AI] for that,” Bogart says. “But I just think there’s an element of AI that doesn’t understand how it’s going to make people feel. It doesn’t think in that way. To write music that stands the test of time, I don’t think AI can do that yet. [Until] we get to the point where AI is completely sentient like the movies and you can’t tell who’s a robot, I don’t think anyone who’s creating music that makes people feel something has to worry.”