In 2023, when the Screen Actors Guild–American Federation of Television and Radio Artists (SAG-AFTRA) strikes shut down Hollywood for four months, seeking protection from, among other things, artificial intelligence (AI), many in India wondered why nothing of the sort was happening there. One of the concerns of the movement was how big studios were going to use AI to replicate likenesses of artists in ways that could be exploitative. The strike ended in a three-year agreement that guaranteed fair pay and included provisions that required performers to approve how their voices were used.
Nothing of the sort has happened in India since. But there’s a small, niche section of the entertainment industry where the effects of AI are being intensely felt: the dubbing and voiceover sector. Gigs have already started to disappear. As for the nature of work, these are strange times in the Indian industry. If you are a dubbing artist, you may find out that your voice has been used in a film you’ve never worked on.
Why the voice sector? Perhaps because some of the advancements in generative AI — like the text-to-speech model or voice cloning — have brought into focus aspects like copyright, compensation and consent in an informal industry of about 20,000 freelancers. The Association of Voice Artists of India (AVA) is more active than ever, issuing circulars and organizing forums for artists in their quest to spread awareness about how to survive and demand fair wages in an environment of uncertainty.
AI isn’t so evolved yet to replace traditional dubbing altogether — it often can’t do emotions convincingly. What it can do is plainer narration, devoid of drama and nuance — perfectly functional for infomercials and corporate audiovisuals, user manuals and even TV promos. Many dubbing artists moonlight in these areas to supplement income — and this sub-sector of the business has taken a major hit. Text-to-speech has cut down costs (for the client) but eliminated the voice artist completely.
“If earlier a voice actor was doing around 15-20 projects a month, now it has come down to maybe six or seven,” says Amarinder Singh Sodhi, the General Secretary of AVA — and also the Hindi voice of Hawkeye (Avengers Assemble) and Blade (Deadpool & Wolverine). “And he or she is probably 40 or 50-plus,” he adds. “It’s not easy to change your profession overnight at that age.”
In 2023, when the SAG-AFTRA strikes were on, actor Joseph Gordon-Levitt wrote a passionate, clarion call of an op-ed demanding fair wages from “tech giants, entertainment giants and every other profit-hungry giant” for using people as resources to train AI. “What’s behind the curtain of AI? The cost of the human labour it took to produce the training data,” he wrote. “‘Generative AI’ cannot generate anything at all without first being trained on massive troves of data it then recombines. Who produces that training data? People do.”
Sodhi and his colleagues are fighting the same fight. Before taking up a project, they want to know what they are getting into, so they aren’t exploited. Tech firms — normally U.S.-based — often approach voice artists to lend their voice to a project in vague terms, without specifying how, where and in what form their voice is going to be used. The lack of laws for AI use has meant that voice artists, particularly the new and less experienced, remain vulnerable to such offers. So, the AVA has taken it upon itself to educate. “If at all you’re going ahead with it, you must understand that your voice is your intellectual property. And it could be misused. And the usage of the voice could be to that extent where your future prospects as a voice artist can be jeopardized,” says Sodhi.
“Earlier, we used to go to the studio, where we were handed over a particular script, and we used to record and then go back home, without ever asking questions. But now the scenario has changed,” he adds. “Our basic rule is that before even going for an audition, ask questions — ask them what exactly you are taking this audition for because random scripts are going to be a complete no-no from here on.”
Aditya Mathur, the channel voice of Nickelodeon, echoes Sodhi. “I can voice in English and then the same can be captured and utilized in numerous languages that AI can do, which basically means I am supposed to be paid for those multiple languages,” he says.
It’s not just about compensation — it’s about consent too. “Tomorrow if someone sends me a clip of a hate speech with my voice in it, I will be horrified, because I don’t support it. And voice is a part of our personality. It’s our identity,” says Rakhee Sharma, who’s been working in the voice industry since she was a child and has voiced Kate Winslet in the Hindi dub of Avatar: The Way of Water, among others. “There are so many ethical and moral issues related to the same,” she adds.
The latest miracle in AI voicing is cloning — a sort of speech equivalent of a visual deep fake. It allows the user to apply the texture of one voice actor to the performance of another. This has made possible a previously unthinkable feat: you can now hear your favorite Hindi-speaking Bollywood star blurt out lines in, let’s say, Telugu, in his or her own voice (and not the dubbing artist’s). This is obviously an improvement on traditional dubbing, giving it authenticity. But where does it leave the dubbing artist? They are still needed, producing the all-important ‘performance’ on which the texture of the star’s voice will be applied.
Dubbing producer Rajashrie Sharma speculates that this could lead to clients paying less. “If a dubbing artist were to charge about ₹3 lakhs or ₹4 lakhs (around $3,500 to $4,500) to dub a theatrical film earlier, there could be an argument now that, ‘We’re not going to air your voice. We’re going to just use your performance and duplicate it,” she says.
Just how popular a tool is voice cloning likely to become in films and streaming series? It has already been seen in films such as Kalki 2898 AD and Vettaiyan — both South Indian films with a pan-Indian appeal. And Kannada filmmaker and founder of the voice cloning studio AI Samhitha, M. G. Srinivas, says that “most of the big films which are releasing in multiple languages are doing it right now.”
Srinivas himself has used voice cloning in his last film, Ghost, and is excited about its prospects in Indian cinema. He says voice cloning is not going to mean underpayment of artists. “It’s not like anyone — any average voice artist — can come and give the voice, and we can clone it into the actor’s voice,” he insists. “It will not work like that. The dubbing artist must be very skilled. He has to do the right modulations. He must be technically strong, wherein the bass voice has to be a certain way and so on.”
Srinivas, who used to be a radio DJ, says that he knows the importance of voice. “Dubbing artists bring something that is original, authentic. Currently, there is no technology to replace it, but five years down the line, who knows?”
The way ahead, then, for voice artists, is to adapt to the relentless march of technology. But insiders say there needs to be regulations in place. Unlike in the U.S., where clauses and contracts have been signed pertaining to the use of AI, in India there have been no binding industry agreements or intervention from the government yet. “Any government or country would want to leverage AI. It is such a technology that it can boost productivity, but it can also disrupt a lot of lives,” says Sodhi. His colleague Ankur Javeri — who’s voiced all of Virat Kohli’s ads until recently and is the Hindi voice of Goku in Dragonball Z — says that to have that kind of impact, other unions from other post-production units must join hands too.
Says Javeri: “The SAG-AFTRA people met a lot of resistance from the big companies, be it 20th Century Fox or Paramount, but eventually they had to buckle under collective bargaining. We need something like that.”