close
close
Creating songs with AI is a blast, but also uncomfortable

Creating songs with AI is a blast, but also uncomfortable

Creating songs with AI is a blast, but also uncomfortable

SAN FRANCISCO — Fun fact: The closest thing this newspaper has to a theme song is this John Philip Sousa march you’ve definitely heard before. It’s a classic, for sure, but perhaps we can do better.

Sadly, I’m no songwriter — so I turned to AI.

This week, Suno, an artificial intelligence start-up that lets you create songs by plugging in just a bit of starter text, released an iOS version of its app. In doing so, Suno arguably made it easier than ever for regular folks like you and me to whip up music on the fly.

That probably wasn’t welcome news to the handful of record companies that sued Suno in late June, arguing that the company’s tool can only generate tunes because it chewed on untold numbers of their copyright songs to learn how. (Suno, for one, has said its technology is “transformative.”) Still, the app remains live and free to download — for now, anyway.

And since the app dropped a few days ago, what started as a silly experiment to generate catchy, journalism-themed tunes has turned into a minor obsession for me. As it turns out, creating full-blown songs on a whim using AI is genuinely a blast, but it also began to reshape my relationship with music in ways I didn’t feel great about.

Here’s what Suno can do and why I felt a little unnerved after living with it.

Getting started with Suno is simple: Just create an account, decide if you want to pay extra to create more songs each day, then start plugging in 200-character prompts.

Generating those songs can take from seconds to minutes, depending on whether you’ve paid for a higher tier of service, and your requests will always generate two tracks for you to review.

Your musical tastes are probably different from mine, but I already knew what I wanted my first attempt at a new Washington Post theme to sound like. Bright, jangly guitars were a must, as were meandering, adventurous bass lines and journalistic lyrics.

But when I asked Suno to create just that, he produced a pair of generic pop-funk tracks that used the words “bright and jangly” as lyrics rather than instructions.

GET CAUGHT UP

Stories to keep you informed

(Listen for yourself: Washington Funk 1and Washington Funk 2.)

Maybe this genre wasn’t the right fit. Next up, I fed Suno the following prompt to see if it would copy a specific artist: “early 2000s Paramore-style pop punk, high energy, female vocals, lyrics about The Washington Post.”

Neither of the resulting tracks immediately felt like Paramore pastiches to me, but that might be because Suno completely ignored my request for female vocals. Still, the songs felt like something I would’ve listened to in high school and featured a surprisingly earworm-y chorus:

Telling tales that we need to know

From the city to the world and back

On its pages no turning back”

(Listen for yourself: Postmortem 1and Postmortem 2)

I wanted to keep those lyrics (plus a few tweaks) for my final attempt, so I opened Suno’s “Custom” mode and pasted them back in for another go-round. (Interestingly, if you want Suno to build a song around a full set of lyrics, its website reminds you to only use AI-generated lyrics; the app doesn’t bother to mention that.)

Now, for the rest of the instructions. Going further afield felt like the right move, so I requested that the style of music include the following elements: “j-pop, math rock, female singer, anime theme, instrumental intro, guitar solo outro.”

And for the first time, Suno’s results felt like they fully embodied what I gave it at the prompt — except when both of the tracks abruptly ended, went quiet for a while, and started up the fake guitars again for one last run-through.

(Listen for yourself: Washington! Post!! OP1and Washington! Post!! OP2)

Okay, fine, none of these will ever really replace The Washington Post March — but if any of them had a chance, it’s Postamore 2.

After I finished my AI journalism song spree, I found myself just messing around with Suno, creating dumb little songs with nonsense lyrics and trying to re-create the styles of one-off tracks I loved.

But it didn’t take long before I felt like I was using — and sharing the results — a bit too much. My wife was having a rough day, so I sent her a lovey-dovey AI song, including our dumb pet names, to cheer her up. I cooked up some truly awful rap lyrics and sent a friend four Suno songs built around them in a row.

Then it hit me — I could easily see myself continuing to dash off songs and send them to people as cavalierly as I fire off emojis.

Music is a force for good, for pleasure and healing and activism and reflection. Was all this slapdash music generation serving in some way to devalue music in my life?

Max Vehuni, one half of the indie-pop duo Slenderbodies, talked me off that ledge.

“Music is a way for people to express themselves.” he said. “If there’s another way for you to communicate with your wife, I think that’s really cool.”

Vehuni, clearly, is no AI music doomer — he’s experimented with Suno and services like it for personal projects and says he sees incredible potential for AI as a tool to enhance an artist’s writing and production.

He’s also quick to admit that, while Suno is being sued for allegedly using copyright music as training data, that process isn’t entirely different from what humans do.

“Artists are drawing a line, saying ‘Well, I’m okay with artists being influenced by me, humans being influenced by me. But once a computer is influenced by me, that’s not okay,’” he said. “Is that something to agree with or disagree with? “I don’t know.”

But that doesn’t mean there aren’t any other things to worry about. The rest of my lingering unease, for instance, stems from a worry that I’d be screwing the artists I love by generating music that sort of sounds like theirs, but isn’t.

Fortunately, Vehuni said slenderbodies makes most of its money from touring and that the band is lucky enough to have a fan base that would support it through “post-AI music apocalypse.”

Choosing to directly support the artists you care about, in other words, is more important than ever.

Still, he worries about the possibility that record labels could pitch their copyright song catalogues to AI companies in return for access to models that can create synthetic music they wouldn’t have to pay royalties on. Or that streaming services will create and promote their own synthetic artists and pocket the revenue. (He’s not alone in wondering about this, either.)

It’s too early to know how any of this will shake out. Either way, Big Tech, the music industry and the rest of us have no choice but to keep grappling with AI music creeping into our lives.

“We’ve taken it out of the box, and I don’t think we’re ever really putting it back,” Vehuni said.