Skip to Content

DJs play around with the Riffusion software while attendees listen in the Riffusion office in San Francisco on Thursday April 25, 2025. Riffusion, a tech company which uses AI to generate music and lyrics based off the user’s suggestions, hosts a launch party for its latest AI model. Photo: Tâm Vũ for Gazetteer SF/CatchLight Local

Partying with the AI music generators

At a launch party thrown by Riffusion, industry players and techies alike see an AI-heavy vision of the future of music — for better or worse

Curt Cameruci, the EDM-slash-trap music producer better known as Flosstradamus, is hunched over a MacBook Pro in a conference-turned-green room at the San Francisco offices of AI music generator maker Riffusion. He’s flanked on either side by Victor Tapia (the DJ also known as Goshfather) and Max Vehuni of the indie pop duo slenderbodies.

They’re shooting the shit, bouncing ideas off of each other. It feels a bit like what I imagine a studio session looks like, except, instead of fiddling with mixing boards, Cameruci is feeding snippets of music and prompts into Riffusion. (He even busts out Google’s Gemini, at one point, to craft a hyperspecific prompt to feed the AI music maker.)

The mood is one of giddy excitement; everyone is astonished at what Riffusion can achieve. “The way it does deep house and tech house is insane," Vehuni exclaims, raving about the little sonic details it gets just so.

A white Marshall mini-speaker placed on a nearby coffee table blasts a French house-y dance track. “Won’t stop texting me, won’t stop calling me,” a pitched-up voice sings. Tapia recorded the vocal himself and processed it with another AI music app, Suno, and sent it to Cameruci, he tells me as the song plays. After one garbled mix gets generated, someone says to turn down the “weirdness” bar on the app.

“I think I became a bit of a believer today,” Tapia says.

I worry that I’m seeing musicians plan their own obsolescence in real time.

But at Riffusion’s offices in Hayes Valley, not far from the historic War Memorial Opera House and Davies Symphony Hall, believers abound. The company has invited a mix of Cerebral Valley types, music industry folks, and Riffusion super-fans for a pre-launch party celebrating its new AI model called Fuzz. Around 75 people are stealing bites of fries as they chit-chat over EDM-heavy remixes of TikTok-viral songs. In one conference room, a game of Super Smash Bros. is underway. The mix eventually gives way to a DJ, the high-tech fashion designer Francesca Rosella, spinning songs she generated on Riffusion. Her mixing might be bad, she laments, but the AI songs are good.

Artist Francesca Rosella DJs in the Riffusion office in San Francisco on Thursday April 25, 2025. Photo: Tâm Vũ for Gazetteer SF/CatchLight Local

We try to both make it extremely easy to get started and then just have a huge amount of depth to the product once you get into it,” Riffusion’s CEO, Seth Forsgren, tells me in a brief interview during the party last Thursday. 

The pitch for Riffusion is two-fold: AI music generators “democratize” music creation for non-musicians, but they can also be part of a working musician’s toolkit. Riffusion is “dedicated to furthering human creativity” for the masses, but it is also merely a “new instrument,” it touts on its website.  

You type a prompt into Riffusion — “hard rock with powerful riffs and vocals,” goes one suggestion I get on the service — or “compose” a track with your own lyrics (or some “ghostwritten” by AI) with the sonic “vibes” of your choice. It spits out songs that you can then refine or remix. 

“Some people might argue that generating music is technically just computer-based, so it's not art, but there's a lot of value that comes from people who are less skilled at specific things being able to make music,” Vehuni, of stillbodies, told me earlier in the week over the phone. “I do value that perspective.”

Key to all this is the idea that AI music is inevitable — that this Pandora’s box has already been opened and to not use it would be foolhardy. I am skeptical, and I eventually got comfortable enough to bring up my skepticism to everyone I chat with over the evening, knowing full well that I’m being the Debbie Downer of the party. 

Tech workers, of course, are stoked about the product; we’re in so-called Cerebral Valley, after all. One well-coiffed software engineer tells me that he texted his friends personalized songs he made on Riffusion, and that the results were “absolutely unbelievable.” They felt like studio-quality songs, his friends told him. At least one employee makes AI-generated lullabies for their kids.

This may very well be just a fun party trick in the way that those viral Barbie and Studio Ghibli ChatGPT renderings are. But there may be a future where dinner-party playlists and elevator muzak could just be AI music generated in an endless loop, Vehuni told me.

Employees and attendees, including Riffusion power-user Daniel Stilman, play Super Smash Bros. in the Riffusion office in San Francisco on Thursday April 25, 2025. Photo: Tâm Vũ for Gazetteer SF/CatchLight Local

There are only a handful of industry types in attendance, but they’re excited enough about the service to fly into San Francisco for the affair.

Early into the evening, I get to chatting with an L.A.-based songwriter who has penned tracks for Zayn Malik and a handful K-pop idols. We talk about his impressive bona fides, and I don’t really intend to interview him for this story. But he eventually opens up about why he's there: After a period of creative burnout, he tells me, he started using Riffusion’s early models to bounce around some ideas for elements like toplines and hooks.

It reminded me of what Vehuni told me on the phone earlier that week: “This is like just another option in the playbook that's a little bit more instantaneous, especially at the speed at which music's being produced and released today. That speed is paramount.”

It seems like the industry has taken notice, whatever their reasons may be. It’s certainly big business: One recent estimate suggests that the AI music industry could be valued at nearly $20 billion by 2028. The Christian rapper Lecrae and The Chainsmokers are on the company’s advisory board. Lecrae was in the company’s office a few weeks ago, one employee told me. Alex Pall, one half of The Chainsmokers, has evangelized for Riffusion in the press

Still, at the party, no one really talks about AI music the way they do human artists. Lucy Dacus and Chappell Roan come up as a couple of Riffusion engineers’ favorites — ironic given Dacus basically hates AI music and Roan has decried the replacement of human creativity by AI. (Other names that get brought up: Cult-favorite slowcore band Duster, the Atlanta rapper Lil Baby, and Charlie Puth.)

One Riffusion super-user, Daniel Stilman, gets brought up repeatedly with a uniformly hushed reverence. Stilman, who makes AI music under the name Yolkhead, has pushed the boundaries of Riffusion, multiple people explain. He’s the prime example of a non-musician Riffusion user.

Stilman, by day, works as a transcriptionist, but described himself as an AI artist and poet. He was flown in from Tel Aviv for the party.

“People aren’t public about it because of the stigma,” Stilman says about people using services like Riffusion. He likens it to professional artists using Midjourney; these services become more acceptable to use personally and professionally the more they stick around, he argues. Any stigma, he says, limits people’s creativity. Much like Vehuni, he frames services like Riffusion as a matter of access — that making AI music allows for “more voices, more diverse opinions.” 

CEO Seth Forsgen makes an opening speech in the Riffusion office in San Francisco on Thursday April 25, 2025. Photo: Tâm Vũ for Gazetteer SF/CatchLight Local

Before we head back to the party, I ask Forsgren, Riffusion’s CEO, what direction he thinks the music industry will go in five years. 

“Over the past six months, we've seen just a huge amount of those artists who maybe were afraid become curious, and then become interested, and then become proponents, and then become evangelists,” he tells me. “That journey's going to just keep coming, I think.”

He pauses: “We're not always going to get it right every single time. We're going to make mistakes along that journey. That's part of this radical disruption that we're in.”

What, exactly, is that radical disruption? Is it the tectonic shift in the business of making music? Is it the re-framing of what “making music” looks like altogether? Is it that our playlists could, one day, be exclusively AI-generated recreations trained on the entirety of recorded music? 

A panel later that evening provides the clearest explanation to the disruption, when a moderator asks: “What would you say to people who believe that AI music is soulless?”

If you ask people fundamentally why they think that, it's intangible,” replies Stilman, who is one of the panelists. “Is the music good? Did you enjoy it? Basically, I’m in it for the songs.”

It's an expected answer from Stilman, but then he acknowledges something that is very tangible: “I know, for a lot of people, the artistic process is the fundamental part.”

Stay in touch

Sign up for our free newsletter

More from Gazetteer SF

Footage from a Tenderloin nonprofit shows potential abuse of an unhoused person

CCTV footage shows a worker at a GLIDE facility harassing a sleeping figure in 2024, and the organization won’t answer our questions about what happened

May 1, 2025

Local law firm standing up to Trump for principles (and profit)

As a litigation boutique, Keker, Van Nest & Peters has less to lose from taking on the administration, and may even gain some new clients

May 1, 2025

Turtle Tower’s stir fried pho is the unsung menu hero that can cure any hangover

The pro move at this resurrected Northern Vietnamese pioneer involves adding wok hei to your broth

April 30, 2025

The biggest perps in the Embarcadero are dogs at the Ferry Building

The facility has an explicit policy prohibiting dogs, but lawlessness reigns supreme. Here's our investigation

April 29, 2025