Drake, the Weeknd, Beyonce, Jay Z, everyone’s falling prey to AI-generated music that mimics musicians. Surf YouTube to come across plenty of AI-made songs and covers. In fact, there is a waiting list on covers.ai where you can “make any AI celebrity cover song”. It’s the kind of music that falls in a grey area for music labels.
Take the example of the song Heart on My Sleeve by Drake and the Weeknd (first uploaded by TikTok account Ghostwriter977). It uses AI versions of their voices that, to a casual listener, sound like the real thing. It uses generative artificial intelligence technology to create familiar sounds. It’s the kind of technology that can be a disruptor. The same goes for the clip of Ariana Grande singing the Rihanna song Diamonds. Fact: Grande’s voice has been generated by AI, which has in its favour speed and scale, outcompeting human endeavour, even though the quality is poor.
The push for AI music started as a trend on TikTok (remains banned in India and facing regulatory issues around the globe). Users have easy access to AI voice generators to clone celebrity voices, which has happened in the case of Harry Styles getting “featured” on the Taylor Swift song, Cardigan. There is AI Kanye West doing Hey There Delilah.
But Heart on My Sleeve is far more sinister because of the marketing campaign that went behind it. Universal Music Group flagged the content on streaming platforms, citing intellectual property concerns. In a statement, UMG said: “Which side of history all stakeholders in the music ecosystem want to be on: the side of artists, fans and human creative expression, or on the side of deep fakes, fraud and denying artists their due compensation.”
The Recording Industry Association of America (RIAA) is also worried about the rise of AI in the music industry. Last October, it said: “That use is unauthorised and infringes our members’ rights.”
Drake isn’t taking the news comfortably. A video appeared with an AI-generated version of his voice rapping on Ice Spice’s Munch and it has made him post on Instagram on April 14: “This is the final straw AI.” An AI version of his voice has also been added to Cardi B and Megan Thee Stallion’s WAP and Don’t by rapper and songwriter Bryson Tiller.
Perhaps the biggest AI artiste is a version of Kanye West, who does Coldplay and Queen. David Guetta too has taken a slice of the action, creating an AI-Eminem and playing it during his sets. “I made it as a joke. People went nuts,” he later said.
Tools of the AI trade
The year 2003 was an important year in music history. Two platforms were launched. Torrent file-sharing site Pirate Bay allowed anybody to download music without spending a cent. The other was Apple’s iTunes Music Store — now just the iTunes Store— which asked users to pay for music, with most songs costing around 99 cents (iTunes Music Store has been a success, slowing down Pirate Bay). The model that is now popular sits between the two. Users pay a small amount every month/year to listen to limitless music (Apple Music, Spotify and so on) or watch ads and then access music (in case you are using the free version of YouTube). But what iTunes Music Store did was, in a way, a gift for the music industry.“Nobody had ever sold a song for 99 cents,” Steve Jobs has Steven Levy in 2003. Then came the rise of playlists and algorithms. You just don’t know what will play next. And in this sludge, what if an AI-generated track finds a place?
For the time being, fans can separate AI work from the real thing. What about in the long run? Will some artistes become obsolete? Royalty-free music generators can compose a rap beat, a commercial jingle or a background score. And technology will simply get better.
Let’s look at some of the tools connected with AI music. In 2019, ByteDance (which owns TikTok) acquired the AI music platform Jukedeck, which allows users alter music to match videos. Next year, Shutterstock bought “certain assets” of Amper, which is an AI music platform that auto-generated music based on parameters like mood, length, tempo and instrumentation.
There are services like AIVA and Beatoven that give game developers, podcasters and content creators royalty-free musical backdrops. And let’s not forget Endel, a generative music app that creates personalised sound environments to match user activities.
UMG senior vice-president of communications James Murtagh-Hopkins has said: “The training of generative AI using our artistes’ music (which represents both a breach of our agreements and a violation of copyright law) as well as the availability of infringing content created with generative AI on DSPs, begs the question as to which side of history all stakeholders in the music ecosystem want to be on: the side of artistes, fans and human creative expression, or on the side of deep fakes, fraud and denying artistes their due compensation.”
Unconventional use of AI
As UMG and others take their point forward, things are becoming, for lack of a better word, creative. In the UK, AI is being used by musicians differently. There is a new eight-track album called AIsis: The Lost Tapes, a “labour of love” from the indie band Breezer. The group created the original music during lockdown in 2021 and then decided to bring in the voice of Liam Gallagher, who has turned out to be AI technology. Breezer, in a way, began when Chris Woodgates and co-writer Bobby Geraghty started writing songs in 2013. Covid lockdowns made them start the band. To get Liam’s voice, Geraghty took various a cappella recordings of Liam to train his AI model. “Our band sounded exactly like Oasis. So then all I had to do was replace my vocals with Liam’s,” he has told The Guardian. The 33-minute concept album basically reimagines the Oasis sound. Except for the AI voice, the songs are original and the music/lyrics are of Breezer’s.
Courts and regulators have many questions before them about ownership when it comes to AI music. At the moment, protected intellectual property can be created by humans, but the grey area has to do with musicians collaborating with machines. Many years ago, musicians like Tom Waits and Bette Midler argued in court that they had a right to not just their musical compositions or recordings, but their voices. They were in court to fight sound-alike imitators in advertisements. But that was in the US.
What about using generative AI to replicate the voices of dead artistes, like that of Freddie Mercury? The team behind BohemianRhapsod.ai allows users to conduct a choir of 16 AI-generated versions of Freddie Mercury to deliver Bohemian Rhapsody.
This year, Google announced it had created an AI tool called MusicLM that can generate music from text. But the tool hasn’t been released for public use, noting in its paper that about one per cent of the music generated matched existing recordings. It’s not about AI music being bad and human-generated music being good. What all of us want to know is who created the music that we are listening to. There needs to be transparency.