E-Paper
From call-waiting tunes to in-game music, AI is in your ear already
AI has been creating behind little tunes we hear every day. Could it create the next hit? Composers and engineers listen in
Think of all the moments in which music fills our lives without us realising: Call waiting tunes when we’re on hold with customer service, theme songs in video games, audio logos (such as Netflix’s tudum and Intel’s five-note clip). Five years ago, these would have been composed by a human being. Now,they’re more likely to be birthed by AI.
Even a familiar sounding song might have come out of a machine. Heart on My Sleeve, which went viral last year, sounds like it was sung by Drake and The Weeknd. It was actually written and produced by TikTok user ghostwriter977, but made (via AI) to sound like the Canadian artists.
The bots are getting closer and closer to creating music that sounds like the real thing, says Mansoor Rahimat Khan, co-founder of BeatOven, an AI-driven start-up that helps composers and content creators create exclusive, royalty-free music. “Technology has started sounding good,” he admits. “Back in 2019, there were three to five AI music start-ups globally, which had 3 million users. Today there are at 50 to 100 million users who generate soundtracks on AI products.”
You’ve probably heard some of it already. It’s in the background music on YouTube videos or in travel blogs. Albums of low-fi ambient music – stuff designed to play while people shop, ride the elevator, wait in reception areas – are already on Spotify. “AI is taking over the space where you listen to music passively,” says Khan. “In active listening cases, as when Arijit Singh is singing, it hasn’t.” He estimates that in by 2026, AI will compose most passive-listening music, leaving human composers to create the more meaningful work.
Production houses in India already use AI for basic audio tasks such as dubbing and same-language subtitles. Mayur Jumani, 31, music composer and content creator from Mumbai says that apps such as Suno AI, create good quality tracks via prompts. Jumani has experimented with Izotope Ozone, which is designed to help with mixing music. He’s also used Soundverse.AI, via which he made some pop music samples after offers prompts about the genre, the instruments he’d like to hear, the tempo, and the melody. Those short audio clips ended up in songs for his Instagram videos. Of course, he could have come up with it without AI, but it would have taken a lot more effort and a bit more time; longer than the AI’s 15 minutes for sure.
It’s made real-life musicians worry. Last month, Billie Eilish, Nicki Minaj, Katy Perry and the estate of Frank Sinatra were among 200 artists who signed an open letter warning against the “predatory” use of artificial intelligence (AI) in the music industry. Among their fears are a dip in overall music quality, the likelihood of bots stealing artists’ voices and likenesses, and artists being denied fair compensation for their work.
Those at BeatOven know exactly how challenging it is for a machine to perform a creative job. Most machine-learning models follow the same rough formula: They comb through massive amounts of data (lakhs of books, thousands of paintings, decades worth of movie posters and such) to create new work based on prompts. “AI such as ChatGPT have scraped the internet to build those models. You can’t do that with music,” says Khan. “To build data for music models, you need consent from the artist. You can’t emulate any popular person’s voice without their permission.”
It means that a bot can’t yet compose the next Billboard charttopper. Heart on My Sleeve hit 15 million views on TikTok before it was taken down by Universal Music Group, which represents Drake and the Weeknd. It was even submitted for a Grammy award for best rap song and song of the year – prizes that usually honour songwriters. But to any human ear, the sound seems eerily repetitive.
Meanwhile, at BeatOven. Khan has signed up with local musicians such as Bengaluru band Parvaaz and Asad Khan from AR Rahman’s line-up, to use the musicians’ songs to create AI music content. It took a lot of convincing. “But if artists are getting paid for it, especially recurring revenue, it makes sense,” says Khan.
Delhi singer-songwriter Aronjoy Das, 26, recently used Udio, an AI-based music-generating programme, to create a series of songs. He gave it vocal prompts about the genre, put in lyrics and picked instruments. The software merely added the tunes. “Once people can easily create their own songs based on ideas they have and sounds they prefer, it will become another form of self-expression,” Das says. “More people will be able to simply give instructions to the AI and create a song without learning how to play an instrument. Besides, it’s much quicker to get things done: you don’t have to record every instrument individually and then mix it.” He’s pumped about the idea.
