Recently, music composer A.R. Rahman tweeted that the song Thimiri Yezhuda from the Tamil film Lal Salaam uses an Artificial Intelligence (AI) software to recreate the voices of singers Bamba Bakya and Shahul Hameed. Bakya died in September 2022 at the age of 42 and has sung songs in films including 2.0 and Bigil. Hameed, who died in 1998, worked extensively on movies such as Gentleman and Kadhalan. Though their voices were used with the permission of their families, who were also compensated, the move has raised a debate among artists about the use of AI. Is it ethical to use AI to clone voices for creative purposes? Sound engineer and music producer Sai Shravanam and playback singer Haricharan Seshadri discuss the question with Srinivasa Ramanujam. Edited excerpts:
As musicians, what were your first reactions when you heard the song?
Sai Shravanam: I was extremely happy to hear the voices of Hameed and Bakya. I was also happy to see that their families were remunerated for the song. A lot of artistes are like comets – they come, deliver big hits, and then vanish. Both these artistes are physically not here now and bringing their voices back brings back incredible memories. A.R. Rahman gave us Roja, which shook every boombox in the country, and now he is bringing the latest technology to the audience. He is timeless, and he has now used ‘Timeless Voices’ to give us this. I am in awe of this technology that brought someone “back to life”. For me, it’s a beautiful feeling.
Haricharan Seshadri: When I first saw their names, I thought it was a mistake. Then, I saw the news and heard the song. I was overwhelmed and was curious about how this process happened. It was bound to happen sometime in the music industry. But it also made me think about my future as a singer.
Is it ethical to use AI to clone voices for creative purposes?
Sai Shravanam: Ethics is personal. AI [in this case] is [doing] nothing but mimicking human talent. An atom bomb, for instance, is an incredible invention, but it has been put to bad use. The invention is not the problem; it’s the issue with the way human beings use it. Even without AI, you may hear a song somewhere, recreate it, and make billions of dollars with it. Ethically, you know it’s not your work and that is also wrong. But AI can never replace human singers and the output that is the result of a creative process.
Haricharan Seshadri: We have been using technology for a long time now to produce songs and make them sound good. Even the auto-tuner that is common in the music industry is a form of AI: it corrects notes based on the inputs given. A process to legalise the use of voices and safeguard the intellectual property of every person and voice needs to be in place. I feel the use of AI voices can be governed. The other aspect revolves around permissions. If a person is alive and approachable, an AI tool need not be used, but if the person has passed on and their voice needs to be recreated, their family and friends need to be aware that such a thing is happening, and also be properly compensated. This is important, especially in a country like India where policies are not enforced. As a singer, I am still fighting for royalty rights for songs I have sung.
Can you elaborate on some instances where technology and AI tools have helped you in your line of work?
Sai Shravanam: AI has been a very important part of my journey. You might have seen documentaries talking about how auto-tune helps people who cannot sing, but you will have to give something for auto-tune to work. In Indian classical music, you have ragas like Thodi and Begada which have micro notes. In those cases, auto-tune sometimes helps, but sometimes it does not. As a musician-cum-engineer, when I know how to use the tool correctly and feed the right settings (into the music processor), I have been successful in using auto-tune even in a raga like Miyan Malhar or Thodi. It is up to us engineers to use it properly.
These days, I sometimes get 200 tracks of audio, with many musicians recording them in different places. They are of different quality – some are very loud and some very quiet. It is overwhelming to hear 200 tracks that are not standardised. If we put them all into the plug-in, it levels everything so that I have an optimal mix that my session requires. AI saves me at least seven hours of work, and I can concentrate on creativity. Today, we have smart dynamic processors and that helps us save time. I f I didn’t have those tools, I don’t have the luxury of delivering an incredibly sounding master within the deadline. AI as a tool has helped me greatly in areas that are not creative-driven; it has helped me in mundane things.
Haricharan Seshadri: AI does not help only in music. Recently, a presenter used ChatGPT to come up with 75 questions to pose to me. I was overwhelmed. Though I am a musician and singer, I have to sometimes record myself at home and send it to a composer. I know how to sing and how good a take is, but there are other aspects in a project, such as editing and fading in and so on, that a sound engineer usually does professionally. AI helps me do all that. I have even tried training my own AI voice model and used it to record a song, and then compared it with my original style of singing. But a real singer cannot be replaced with AI, because we add bhaavam or feeling to a song. I fear that that the increased use of AI tools would result in mediocre music.
Do you feel it may disrupt the progress of creativity and humaneness. And if yes, how do we tackle that?
Sai Shravanam: Creativity is God’s gift. It doesn’t come from you but rather through you. Art is not about learning, but rather about exploration. AI mixing and mastering tools are here, but human interactions and emotions can change everything. To me, AI can help in mundane things but never in the creative process. It can never replace a Haricharan walking into my studio, learning the lines, and delivering them.
Haricharan Seshadri: It will disrupt creativity, but only for a short while. People will get tired of it soon. From a film industry perspective, a lot of mediocrity is glorified because of reels and social media views. Lyrics are not meaningful but rather just based on some rhyme scheme, just to make it trend. AI will have its day. It will be glorified and we will grow tired of it.
In this marriage of music and AI, where is the future headed?
Sai Shravanam: There is responsibility among human beings to use AI the right way. We must understand that arts or music is not a product. It has unfortunately become a product. The human brain is about perception, and what I hear today as a sound engineer is not what I hear tomorrow. A computer can only give one output, and we should realise that arts is not just a money-making industry.
Haricharan Seshadri: There needs to be an AI Ethical Usage Board in every industry. Deepfake videos are also posing a threat, so there needs to be some kind of regulation. All intellectual property needs to be registered. Tomorrow, someone might be able to use my voice for a song without permission and I should be able to sue them through proper legal channels.
Listen to the Parley podcast here.
