Almost live music: How artificial intelligence is taking over the charts

The band The Velvet Sundown has suddenly become popular, at the time of writing they have almost a million regular listeners and some songs have racked up over 2 million listens on Spotify. She performs a mix of indie and pop rock, and the photo from the "concert" shows the four performers, in classic images of rock musicians from the '70s, singing inspirationally from the stage. The only nuance is that everything from the music to the band photos was created by AI. How are AI tracks, artists and apps changing the landscape of the music industry, what's in store for labels and artists, and how can they capitalize on it?

The Velvet Sundown case caused a lot of media buzz, with The Guardian, CNBC and the UK's The Standard writing about the band. A conversation between a certain Andrew Frelon and The Rolling Stone magazine in early July, where he called himself a spokesman for the non-existent band, also added to the hype. That said, the band's official X social media account denied any connection to Frelon. "We understand the intrigue of our project, and we're not here to dispel the mystery," was the comment the publication received from the band's Spotify account. The hoax only fueled the public's interest in the new phenomenon.

Right now, the "Biography" section of the band's Spotify account states rather vaguely, "is a synthetic music project created under human creative direction and composed, voiced and visualized with the support of artificial intelligence." "It's not a gimmick - it's a mirror. An ongoing artistic provocation designed to challenge the boundaries of authorship, identity, and the future of music itself in the age of artificial intelligence," the section reads.

But even though the band doesn't exist, that hasn't stopped it from making real money; according to estimates from streaming royalty calculator ChartMasters, The Velvet Sundown song has earned about $34,235 in 30 days across all streaming audio platforms.

But the story of The Velvet Sundown is far from the first such case.

Music Business WorldWide columnist Tim Ingham also found "dark country" collective Aventhis with more than a million monthly listens on Spotify, and the algorithms have started giving it other artists as well. "The Devil Inside is a 'band' that has about 700k monthly listeners on Spotify. Like Aventhis, they are 'verified' by Spotify, and their most popular track, Bones In The River, has 1.6 million listens to date," he wrote.

The generated tracks have already been hitting European charts and TikTok trends, and world-renowned producer Timbaland recently launched Stage Zero, a project that uses AI to create music and artists, and the first of these is TaTa.

Timbaland calls the new genre A-pop, from the acronym AI, meaning AI. "She's not an avatar. She's not a character. TaTa is a living, learning, autonomous musical performer created with artificial intelligence. TaTa is the beginning of something bigger. She is the first artist of a new generation. A-Pop is the next stage of cultural evolution, and TaTa is its first icon," Music Business Worldwide quoted Timbaland as saying.

"If other producers start doing this ... and it's going to create a completely different model for the music industry that we can't predict yet," Jason Palamara, an assistant professor of music technology at the Herron School of Art and Design, tells CNBC.

At the same time, the number of generated tracks is growing exponentially. In January, French music streaming service Deezer introduced an AI-based music discovery tool; in April, it reported that about 18% of all tracks uploaded to the platform are entirely generated by artificial intelligence. Spotify does not disclose this data.

At the same time, AI-generated tracks are not labeled (with rare exceptions) and every year it becomes more and more difficult to distinguish them from human-generated tracks. A new image of the music industry is forming before our eyes - with new players and rules that are yet to be established.

Now literally anyone can become a music AI producer and musician. To create your own hit on the basis of a short promt you will need literally a couple of minutes, the simplest compositions can be created for free.

With an $8 or $24 per month subscription on Suno and Udio, you can access additional paid features - like creating four projects at once. There's also the option to upload your own lyrics or a half-finished track to work with in the app, or assemble a band of AI session musicians.

The Magnificent Seven companies have not been left out of the explosion of AI music and are releasing applications and tools for its generation. Meta has released JASCO, AudioCraft and MusicGen, Google has MusicLM, Lyria and Music AI Sandbox, and Apple has introduced AI solutions to its proprietary music creation program Logic Pro in Spring 2024. Microsoft decided to bet on the Suno app and included it as a plug-in in Microsoft Copilot.

Tracks by The Velvet Sundown, TaTa, The Devil Inside and Aventhis were all created using the Suno app. In an interview with CNBC, Jason Palamara calls the two apps - Suno and Udio - the "gold standard" of modern AI-based music generators. They do not require any specialized knowledge from the user and are able to create a track based on a single promt in a few seconds.

Suno became widely available starting Dec. 20, 2023, after launching a web app and partnering with Microsoft, the company said on its website.

Suno was founded by Michael Shulman, Georg Cusco, Martin Camacho and Keenan Freiberg. They all worked at artificial intelligence startup Kensho before founding their own company in Cambridge, Massachusetts. According to Clay, the company has raised $125 million in a second round of investment and is valued at $500 million. Crunchbase lists Lightspeed Venture Partners as the lead investor.

In February, Amazonunveiled a new version of its artificial intelligence-powered Alexa voice assistant, and among the next-generation Alexa+ features is integration with Suno. In July, it was revealed that Suno had hired Paul Sinclair, former CEO of Atlantic Records, as chief music officer to oversee the integration of AI tools into the songwriting process.

The second notorious app is called Udio. It was founded by Yaroslav Ganin and David Fannin Ding. Last year, Udio received $10 million from a group of investors, including Andreessen Horowitz, Bloomberg wrote.

Udio's seed round has attracted high-profile tech and music investors, including Instagram co-founder Mike Krieger, Google DeepMind executive Oriol Viñals, musicians will.i.am and Common, producer Kevin Wall and music distribution platform UnitedMasters. The scheme of work is the same, so is the price of a paid subscription.

The introduction of AI into the music industry is generating revenue for some and losses for others.

According to a 2024 report from CISAC, AI service providers' revenues will grow significantly over the next five years, while content creators are at risk of losing a significant portion of their current revenue. Music and audiovisual content creators face the risk of losing nearly a quarter of their revenue by 2028. This translates into a combined loss of €22 billion over the five-year period (€10 billion in music and €12 billion in the audiovisual sector).

Artists and recording studios, with varying degrees of success, are already fighting for fair use of their work and the introduction of new royalty schemes.

In June, representatives of the British music industry - from Elton John to Dua Lipa - protested against the Data Use Bill, which makes it easier for big tech companies to access artists' copyrighted material. The protest didn't work, and the bill was passed in June, putting musicians in a vulnerable position, Standart points out.

The big license holders are doing much better at getting fair copyrights.

Last June, Sony Music Entertainment, UMG Recordings, part of Universal Music, and Warner Records sued Suno and Udio for copyright infringement in model training. Some of the songs generated were virtually identical to the original tracks on which the models were trained. The Recording Industry Association of America called model training willful copyright infringement "on an unimaginable scale" and demanded payment of up to $150,000 for each work used, which could total billions of dollars.

Nine months later, the same companies are at the negotiating table. Now they're discussing AI music licensing agreements that would allow startups to continue to learn from catalogs - as long as the labels, and eventually their artists, get paid.

According to The Wall Street Journal, the record companies are demanding license fees, compensation for past use and minority stakes in both companies.

To determine how much money artists and labels should receive, the companies want startups to develop identification and attribution technology (similar to YouTube's Content ID) to track when and how a song is used, sources said. That means companies and artists would be able to get paid for that usage.

In addition, the companies want to be active participants in the music products produced by AI startups, including having a say in how they are designed and how they operate, the WSJ writes.

These negotiations aren't just an attempt to settle a lawsuit. It's about establishing new rules for using copyrighted music to train AI.

The battle is for a promising market: analysts from Grand View Research in their report estimated the volume of the AI music market in 2023 at $440 million and predicted that it could grow to $2.7 billion by 2030, it will grow by 30% per year. The Research and Markets portal is even more optimistic: in 2024, it estimated the AI music market at $419 million and predicted that it would grow to $4.3 billion by 2029.

This article was AI-translated and verified by a human editor