AI-Music Arms Race: Meet Udio, the other ChatGPT for music

Just last year, Many experts believed that an AI model capable of producing full, high-fidelity songs from text messages would not be coming anytime soon, but now, an arms race is underway between competing music creation models that do just that. SunoThe company's v3 model, released to the public a few weeks ago, was a remarkable breakthrough, particularly in realistic human-sounding vocals — and today, a formidable new competitor arrives via the newly launched company Audio. The two companies' performance seems very comparable, although some early adopters have suggested that, on average, the Udio's performance may sound crisper than the Suno's, with less sonic ambiguity that can betray the provenance of the tracks created by machine.

Udio's product has come together remarkably quickly since it was founded last December by four former researchers at Google's AI research arm, DeepMind. They are backed by a number of tech heavyweights, including 16z (aka Andreesen Horowitz) and Instagram co-founder and CTO Mike Krieger. “We've had really good support from the day we invested,” says one of the company's co-founders, Andrew Sanchez. “So the technical co-founders were kind of able to hit the ground running because we could do it all very quickly.”

There are also some notable music names on Udio's list of early investors, including Common, producer Tay Keith, United Masters vet Steve Stoute, and Will.i.am. In a press release, Will.i.am, who has long been an evangelist of AI's musical capabilities, was tight-lipped about the company's product: “This is a brand new Renaissance, and Udio is the tool for its creativity. of the time,” said the artist. , which was consulted during product development. “With Udio you can create songs through artificial intelligence and your imagination.”

While neither company will confirm or deny it, there is substantial reason to believe that both Udio and Suno were trained in copyrighted music without permission, a practice recently condemned in an open letter by Artists Rights Alliance, signed by Stevie Wonder artists. to Billie Eilish. The group reiterated its stance in a new statement to Rolling rock. “Using artists' work without consent, credit and compensation is not only unethical and irresponsible and destructive to cultural gifts, but also illegal,” said Jen Jacobsen, Artist Rights Alliance Executive Director. (The question of whether copyrighted material can legally be used to train AI is currently being decided in various court cases.) “The Alliance for Artist Rights urges AI platforms to back down from this reckless path and, if they don't, we urge rights holders to take immediate legal action to stop them.”

In Rolling rockUsing a pre-release beta version of Udio, the service produced two separate ones songs on a day when he was visibly using a voice indistinguishable from that of the late Tom Petty. “We have a filter to detect cases where the voice sounds too similar to some existing artists,” says co-founder David Ding. “Of course, this potion is not perfect. It's also an AI system, and we'll keep iterating on it. But it's definitely not meant to be.” However, Udio may have future plans to offer voice clones more formally — the company's press release mentions an upcoming plan to give artists “financial control over their voice similes.”

Unlike Suno, which is focused on putting music-making tools in the hands of average consumers, Udio also sees itself as a tool for musicians, and the founders say its creative abilities along those lines should ease concerns. of the authors regarding the use of the training data. . “We've been driven by music people from the ground up,” says Sanchez, “and that means building a product that will allow those people to make great music and, to be clear, make money. of this music in the future”. Sanchez emphasizes his belief that the company's output is “transformational”—that, in other words, the model makes something new out of its training data, an argument that ChatGPT is also using in its current court cases over the use of material that is protected by copyright.

An AI blues song that Rolling rock created with Suno's v3 went viral last month. above is a competing effort from Udio. We used the prompt “solo acoustic Mississippi Delta Blues for a sad AI” — like Suno, Udio calls OpenAI's ChatGPT via API to write lyrics unless you provide some of your own. We also asked Suno and Udio to produce diss tracks for each other. here are those results.

Suno currently produces two-minute music clips from her commands. Udio is more customizable but perhaps less intuitive to use, starting with 30-second segments that can be expanded to user specifications. There are even more detailed controls that the company already offers privately to musicians, and co-founder Sanchez says any creator with concerns about the company should reach out to them. “We're open for business,” he says. “Call us. We'd love to talk and get you there and then you'll see, “Oh, wow. This is really cool. And I'm less afraid of it now because I feel like I own it.”