AI text-to-music tools are thriving. Recently, Suno raised $125 million in investments (the company would not disclose details of its financing), puzzling the entire industry: if their technology is already so impressive, what can we expect from a company with such backing? Perhaps we'll even hear Suno's tracks in new seasons of House of the Dragon if rightsholders don't nab them first.
Many have joked that this money will come in handy for covering legal costs due to lawsuits from rightsholders of the songs used to train Suno's models. Ed Newton Rex suggested in his Substack post that the huge investment is due to investors' FOMO: they missed out on OpenAI, so they're eager to grab a piece of this pie instead. Incidentally, Ed Newton Rex was one of the many who speculated that Suno used copyright-protected songs to create their product, and one of the few who is closest to proving it.
In a column for Music Business Worldwide, Ed wrote about how entering specific prompts into the Suno app produced compositions suspiciously similar to copyrighted songs.
Many other creators seem to agree.
Why listen to my music when AI can give you this beautiful magnum opus? pic.twitter.com/Yk8VexASdT
— Benn Jordan (@bennjordan) April 11, 2024
Some time later, Ed wrote a similar column, but the main subject then was Udio—the closest competitor to Suno in the AI music generation. Users and AI music enthusiasts argue over which product is better; some say that "Udio is 6-12 months ahead of Suno, in AI time." This may well be true, as Udio, like Suno, also recently raised $10 million in investments, attracting a constellation of star-studded investors.
With backing from Andreessen Horowitz (the venture capital titan that argues AI models should gobble up copyrighted music freely), and an eclectic mix of investors: from artists like Will.i.am and Common to the co-founder of Instagram, Mike Krieger, Udio is making waves. The latter believes Udio could do for music creation what Instagram did for photography. Krieger is quoted in the announcement release as saying “just like Instagram brought photography sharing to the masses, I believe Udio has the power to bring music creation to the masses as well”. One can almost hear the groans of musicians as they brace for a tsunami of AI-generated tracks flooding streaming services.
Udio users are surely revelling, churning out 1,200 free songs a month. It starts with text prompts: you type in a description of a music genre, topic, lyrics, and (as the company’s press release oh-so-eloquently puts it: “artists that inspire”), and voilà—a piece of music materialises within 40 seconds. This creation can then be “remixed” through further text prompts until the user achieves their desired masterpiece. The solution even boasts the ability to create 15-minute tracks, overshadowing Suno's mere 4-minute limit.
Udio’s CEO told Music Ally that developing Udio has been a balancing act between making its service as accessible as possible for casual users, while also being powerful and flexible enough for professional musicians and producers. Udio’s ambition stretches beyond merely being a music creation tool, though; it wants to build a community around it—a strategy we’ve seen before with the likes of BandLab and Splash.
But here's where it gets murky. Udio, much like Suno, hasn’t disclosed the specifics of its training data. In an interview with Music Ally, two of Udio’s founders answered a direct question on this without really giving any details: they said they “train on a large amount of publicly-available and high-quality music”, they “create transformative new music”, and they “have very strong artist filters… to make sure we don’t regurgitate anything [copyrighted]”.
Yet, Ed Newton Rex's experiments with Udio have resulted in music eerily reminiscent of famous hits. As the line between inspiration and imitation becomes even more blurry thanks to AI, the question of how to fairly compensate original artists looms larger than ever.
“The goal here, obviously, is not to reproduce the Beatles or stuff like that. If I want to reproduce the Beatles, I could just listen to the Beatles! It’s to understand the fundamental underpinnings of music, and take this knowledge and use it to create transformative new music that’s inspired by existing music, but is completely novel,” Udio founders say in their Music Ally's interview.
Udio’s press release boasts that the company is “in discussions with a number of artists who want to leverage AI in their workflows and find new ways to make money through its tech.” They claim to have “spent time building its business model to be beneficial to artists,” which includes giving artists financial control over their voice likeness and ensuring generated tracks do not infringe on copyright.
In the first two weeks after Udio’s launch, over 600,000 people jumped on the bandwagon, according to co-founder Andrew Sanchez, speaking to Bloomberg. Users have been churning out an average of 10 songs per second. That's 864,000 tracks a day—more music than anyone could possibly listen to. With such an overwhelming output, one has to wonder how many of these tracks are making their way onto streaming platforms and what this means for the royalties pie that artists and rightsholders depend on.
Another big question is whether there will ever be an industry-standard technology capable of 'fingerprinting' the copyrighted material fed into AI models. Such technology could ensure that creators of original works get their due royalties. But until then, the fate of AI-generated music might just be decided in courtrooms around the world. For now, the music industry watches with a mix of awe and trepidation as AI tools like Udio and Suno continue to grow.
