If you’re a fan of musical classics and wonder how Mohammed Rafi would sing contemporary hits like Shah Rukh Khan’s “Haule Haule” from Rab Ne Bana Di Jodi, you’re in luck. Instagram is a hotbed of retro remixes and regenerations and at the centre of it all is generative Artificial Intelligence.
Recently, we have heard of AI making strides in several fields; but it goes beyond ChatGPT helping you with your essays. Musicians are growing more and more popular by integrating software into music, creating a category of audience who can enjoy songs from before their times, while others can relive the nostalgia of old music.
Anshuman Sharma is one such music producer with over 1,95,000 followers on Instagram. Now working with Salim Suleiman, Sharma first went viral with his “How to make a Ritviz song in 2 minutes”. Since then, he has continued his journey warping songs, from making Charlie Puth sing in Hindi to his most recent viral addition, “If Rab Ne Bana Di Jodi’s Haule Haule was a Mohammed Rafi song”.
But behind these growing trends lurks the question–what does it mean for a vocal artist if the asset that is their voice, can be recreated by AI? And what about the slew of copyright complications that a reckless use of AI creates?
“Musicians can’t be replaced by AI”
Although Sharma has a growing popularity on Instagram, he considers that a side project. The musical regenerations take him 5-10 minutes and he claims it is only a matter of converting one vocal texture to another. However, he does not see AI putting musicians out of business anytime soon.
“If we look at the state of AI and automated music, it’s not that great right now. It cannot compose music like human beings can. Maybe it will get there in a couple of years but right now, it’s great as an additional tool,” Sharma said. Sharma sees AI as a good tool to experiment with. “Someone sings a part of a song and then AI converts that vocal file to another texture. So we cannot say AI is making music right now.”
So how does Sharma make Mohammed Rafi sing contemporary music? Sharma adds that AI has to be fed a song sung exactly in Mohammed Rafi’s style so that it can pick up the right samples. “If you don’t sing exactly like him, then the output won’t be that close.”
However, Sharma is certain that the human spirit in music is irreplaceable. “Maybe in 10 years, AI could compose music. Music is based on a feeling, which you can’t translate into coded language.” He acknowledges the versatility of AI, but the complicated nature of music composing guarantees that it will remain in the hands of human beings for a good time to come.
Sona Mohapatra, singer and songwriter, echoes the same positivity about AI in music. “Whenever new tech comes in, it sends clear ripples. Everyone worries about becoming redundant. I don’t particularly see AI replacing humans in a manner that should worry any of us,” she said. As an independent artist not counting on record labels, she views AI as beneficial to her work. “Creating lyric videos and digital arts affordably while being able to distance one’s self from the commercial world of posturing and numbers, AI is a useful tool.”
Mohapatra agrees that AI could never replace art. According to her, it’s all a game of trends, where there are songs that become hit numbers without even being a conventional song. “Sometimes out-of-tune singers become trendy among kids,” she added. However, the “real deal” as she mentions, is when a singer is able to move the audience with their music. “No AI can replace the feeling of catharsis when you sing along with a performer,” Mohapatra added.
Distancing herself from the conservative views of looking down on retro remixes, she says that no amount of great production, be it by Vocoders or Melodyne, can replace great storytelling of music. “It is all a transition phase and I want to channel it in the best way possible.”
However, Mohapatra grants that boons can quickly become bane if left unmoderated. “It is great to get a framework of ChatGPT but it cannot be allowed to replace human thought. And artists should be aware of what kind of creative help can be taken from software.”
Shubham Kabra, a musician with a steadily growing following on social media says it may be easy to copy someone’s voice, but it depends on what purpose they’re doing it for. “The generation that’s coming up may have become lazy, but music requires humanity and soul which you can never code into AI,” said, Shubham.
According to him, if a composition can be enhanced with AI, then it can be used as a plus one. “The track has to sound good–so to a musician, the technicalities don’t matter as long as the emotions come out in the end.”
Renowned music director Somesh Saha concurs that although AI is a great tool to have fun with, it cannot emulate a personality the same way that voice cloning cannot reproduce a style. “Beyonce’s style is a result of her personality, that AI cannot replicate. Artists are popular not just because of their songs; they’re also individuals that generate intrigue in the listener’s mind,” he explained.
Questions of consent and copyright
Although the reception of AI in music has so far been positive among musicians, there are still debates on how best to address the copyright question.
Sona Mohapatra argues that it’s all well until someone’s talent is being misused or exploited without due credit to commercials. In her words, if something is unique to you, you should be paid for it. She reiterates that if AI is being used to eliminate human creation, to save up on actually hiring Sona Mohapatra or her brand, then that needs thought and regulation.
Somesh also notes that there is no one good benchmark to say what’s going wrong in the industry. “As far as the sanctity of something goes, if a recreation is uploaded online, there should be a clear indication that no revenue should be generated from it.” Or if that’s the case there has to be a robust mechanism to provide royalties to descendants of these artists (perhaps the AI voice emulators can have a digital footprint of sorts that can be tracked). Hence, the mechanism of such indication on the platforms could solve the issue.
Avadhi Joshi, Partner at VNA Partners (head of Media and Entertainment practice), highlights the critical need for creators to secure personality rights when using the voices of artists/ singers like Mohammed Rafi or Kishore Kumar. While legal agreements usually encompass a clause granting authority to use the name, likeness, and voice of the original artist, obtaining consent from the artist remains pivotal, especially in cases where the artist is deceased. In such instances, securing consent from the legal heir becomes imperative.
Joshi emphasizes that incorporating a hook from a renowned singer into AI-generated content doesn't qualify as truly original. The absence of permission from the original singer in coded songs constitutes direct infringement.
The issue stems from the fact that there is no clarity on who owns these. By fact of law, it’s always the producer who owns songs. But what if your AI-generated content has taken music from somewhere else?
Nayantara Malhotra, Founder and Intellectual Property Lawyer at Ostara Legal, argues that when voices are cloned, lots of parties get affected. Number 1 is the copyright holder who holds the right to license dissemination and right of reproduction and to allow someone to make derivative work. Next, it affects the artist whose voice is being used.
The responsibility is of the content creator who is putting it out on social media. “If the creator were to contest that they’re doing it for entertainment and not revenue generation–even that is not permitted. So to see what’s permissible, you have to go back to the Copyright Act,” said Nayantara.
Are Social Media Platforms Liable?
Can the social media platforms really be blamed? Decode spoke to lawyers to get a sense of who shares the liability of reproduction.
Nayantara said, “Platforms are mere intermediaries. Instagram does not have the primary liability to scrub the infringing content.” It is on the onus of the aggrieved party to reach out to Instagram or send a legal notice to tell them that there is an infringement. If Instagram does not comply, that’s when they can be taken to court.
In fact, if a creator holds a trademark that is not registered, the platform can even refuse to comply with the request for the removal of content. “If the platform has a reason to believe, that the original artist’s rights are shaky, they have the free will to not act upon it till there’s a court order. Because they need a public authority’s seal of approval before acting.”
One aspect is crystal clear: hosting platforms bear no responsibility. Their Terms and Conditions hold the uploader accountable, acknowledging the impracticality of scrutinizing AI-generated content amid the daily influx of numerous song uploads.
Apart from the legal rights, there is a scope for the public to get misled. There is a scope of misusing someone’s voice to even put out something controversial. In that case, who saves the original artist from defamation?
Although the grey areas are endless and the lack of legislation adds to the ambiguity, musicians are still hopeful that the audience can tell between real music and computed ones. “There is an India which connects to real artists and beautiful storytellers. Others will jump on a trend when it’s amusing, but it is a passing phase,” Sona Mohapatra concluded.