When AI has descended its claws into all corners of creativity, the music industry is no exception. Recently, Eros Innovation has announced that it is bringing back legendary singer Mohammed Rafi’s voice via AI. They are also launching seven AI-native artistes, such as Jordan, Tanu, Munna, Langda Tyagi, and Mudit. Also, with the pop icon Taylor Swift recently moving to soundmark her voice to protect it from AI and Spotify moving to introduce AI features, the topic has resurfaced among social media.
The debate over AI in music remains unsettled. Excluding the technology now requires a deliberate effort from artistes and everyone else in the creative chain. Collective Artists Network launched India’s first AI rock band, Trilok, with no human musicians. Some companies are building AI music models and employing humans for research. The musicians listen to AI-generated tracks and offer opinions, and that data trains the next iteration of the software. It is conceivable that Gen Alpha will listen to a new genre — AI-generated music.
The note change
The biggest impact has been felt in music production, notes Joel Sakkari, who, apart from making independent music as Sakre, produces music for other artistes and brands. “It has become common for people to approach me with an AI-generated track for film songs, jingles, or ad music and ask me to recreate it — I proofread the chord progressions, create real guitar parts on it, and legitimise that AI composition,” he shares.
In keeping with this trend, Mack Raj, MD of The Bangalore Studio, a film production company which has transitioned into AI-based work using tools like Google’s Flow, Gemini’s Omini, Kling AI and more while developing an in-house LLM (Large Language Model), says that the tech and lower production costs are drawing artistes trying to make it in the industry. But he strongly believes that AI “cannot replace vocals or the soul of an artistes.” He adds, “If you were a lyricist or a singer before, you’d have to invest approximately Rs 50,000 to get a song produced and spend lakhs on music videos. With AI, the audio production takes Rs 10,000 while videos cost around Rs 5,000 per minute,” he says, adding that the cuts come from saving on musicians and equipment.
Spoonful of scepticism
Established musicians like Vasu Dixit, though, are not sold on the idea. Noting how he can spot AI-generated music all the time, especially on restaurant and café playlists, he says, “AI-generated images used to be very slick looking, with rough edges and extra limbs — to musicians, AI music sounds like that. It’s still in a very initial stage,” says Vasu. For him, music is as much about the process as the result and the step that makes a song true.
Musician Ajith K Prakash, who performs as Ajoopan, uses AI to generate videos for his original compositions, as the process is much easier for an independent artiste. He adds, “The complexity of instruments also matters, and it is difficult to capture the intricacies of instruments like the flute or the violin. Virtual Studio Technology (VST) that was released years ago, before AI that would give digital versions of most of the musical instruments. It was released more than 20 years ago. But even after this long, we haven’t been able to get into the complexities of the instruments. It has its own disadvantages. The music it produces won’t look like a human played it.”
Instruments like the yazh, parai, and kudamuzha carry histories, landscapes, and communities within them, comments Sivasubramanian Muthusamy, aka Sisu, a member of the band Uru Paanar. He notes, “I think AI can get very close to the sound (of indigenous instruments), but not necessarily to the experience around it. As someone still learning music, what interests me is often everything around the note itself. How an instrument sits in someone’s hands, how it responds differently every time, how a musician adjusts to it in the moment. Indigenous and old instruments carry traces of the people and places that shaped them. You can sample the sound of a parai, but you can’t easily sample the generations of practice, rituals, and memories that come with it.”
Like AI-generated texts have problems with inconsistencies, AI music has these issues too, as Joel, who has experimented with the technology but does not use it in his own work, elaborates. “The tones of instruments are odd, sounding somewhere between real and electronic; the instruments’ tone, the melody and the voice of the singer change drastically in the middle. It gets rhythm mostly right but really overcomplicates harmonies, so you end up spending more credits to fix these issues,” he says, adding that it’s especially frustrating when novices, believing that AI is the expert, refuse to listen to reason. He recalls, “It had made up a chord that could only be played if all the strings were jumbled and the musician had seven fingers, but the client insisted that I recreate it.”
One area that has remained mostly untouched by AI is performance. As DJ Ashok Nalwade, notes, while AI generated tracks feature in club playlists, the highlight is still a live DJ. “DJing is 70 per cent what you play and 30 per cent how you engage with your crowd by feeling their energy, talking to them, and getting them hyped up. Maybe someday, AI could do that, but today, it doesn’t take away what DJs bring to the table,” he opines.
Preservation or erasure?
People connect to music in different ways. “Some are happy just listening to a track, while others want to know who made it and where it comes from. Neither is wrong. Personally, I enjoy being in a space where music is happening in real time. Watching musicians respond to each other, improvisations, and moments that weren’t planned,” shares Sisu, adding that old and indigenous instruments carry those human stories within them. “It’s also about preserving a way of gathering, learning, and sharing knowledge.”
The question of where AI fits is one musicians are answering differently depending on where they stand in the industry. For Ajoopan, the technology has already changed how he works with clients. Now, AI lets him generate references quickly, before committing full energy to it. He sees AI edging into lower-stakes work, perhaps replacing amateur-level work in advertisements, perhaps allowing brands to buy tools and generate their own music at a fraction of the cost they currently pay producers. That shift is happening. But he draws strong boundaries on replacing the sound of artistes and instruments with AI.
But will AI help preserve musical heritage or flatten traditions? Sisu says, “I think it can do both, depending on how we use it. AI can be useful for documentation, archiving, and preserving knowledge that might otherwise be lost. But reviving an instrument is still a very physical process. It involves making, listening, adjusting, and learning through trial and error. Being part of a band that works with revived instruments has taught me that a lot of learning happens in the real world. What excites me is not just recovering an old sound, but understanding how these instruments can continue to evolve. I also think AI has great potential as a support tool for live music. I can imagine AI agents helping with sound checks, acoustic analysis of venues and understanding audience energy during large concerts. Instead of replacing musicians, these tools could help artists focus more on performing and connecting with people. That’s the kind of AI I find exciting — technology that supports the live music experience rather than trying to recreate it.”