In late April, a company called Voices.com made the kind of routine corporate announcement that often passes without notice. Based in London, Ont., the company said it had acquired a new domain – Voices.ai – and would launch a platform this summer for actors to clone their voices with artificial intelligence and earn royalties whenever a client, be it a commercial director or a software company creating a virtual assistant, is in need of a narrator.
The reaction on Twitter, however, was swift. “WHY WOULD AN ACTOR DO THIS?” voice actor Tara Strong tweeted. Another tweeted at the company: “You’re on the cusp of a mass exodus.” And, indeed, some people said they would shut down their accounts on Voices.com, a large marketplace for (human) actors to connect with businesses seeking voiceovers for ads, animations, audiobooks, video games and other media.
Will AI take over the world? And other questions Canadians are asking Google about the technology
The company put out a statement clarifying that it has never cloned a voice without the owner’s consent – nor would it ever do so – while, on Twitter, replying individually to some of the most aggrieved tweeters.
“The reaction was a little premature,” says David Ciccarelli, founder and chief executive of Voices.com. “I understand where the fear is coming from, but candidly, I feel like it’s unfounded.”
The blowback shows the deep concern and visceral opposition to AI in some corners of the creative world. The quality of AI-generated voices has developed rapidly in the past few months, and some companies can churn out a convincing copy of a voice based on less than five minutes of audio, and for only a few dollars. Advertisers are interested in the potential cost-savings, startups are offering AI voices for corporate and educational training materials, and Apple is already using artificial voices to narrate audiobooks, all of which has sparked worry that the technology will limit work for voice actors.
For those who depend on their voices for their livelihoods, deriving passive income from an AI clone might be appealing. But for others, it represents an insidious creep toward their own demise.
How AI-driven labs will fast-track the future of everything
Ciccarelli, who founded the company in 2005, has had talks with voice-generation companies for the past few years and found the results to be overly robotic, and ultimately, unusable. That’s no longer the case. “It went from laughable to passable,” he says. The quality improvement prompted him to think seriously about how to jump on the trend, and a business model based on consent and fair compensation seemed like an ethical approach. “We want to protect the talent, as it’s protecting our business as well,” Ciccarelli says.
Today, he sees limited applications for AI clones, which still lack the full emotional range of a human voice, along with subtleties such as sarcasm and comedic timing. AI is better suited to informational contexts, such as public service announcements, digital assistants, corporate and educational training material, and perhaps some advertisements.
A big ad campaign could require multiple versions, swapping out one city name for another or tweaking the pitch depending on the target audience. Ciccarelli said one ad the company was involved with last year necessitated some 2,000 variations – all of which had to be recorded by the actor. An AI voice could theoretically do so much faster, and for less money. (High-end commercial productions will not want to cut corners, he believes, and will continue hiring voice actors.)
Some voice artists Ciccarelli has spoken with are intrigued by the opportunity. The jobs where AI can be used are not necessarily the most fulfilling, either. “Maybe the job’s just not really interesting,” he says. “It’s just a pretty boring corporate training video.”
But that training video is still a paying gig, a niche some people specialize in, and there is pride attached to what others might dismiss as boring work. “One of my first mentors in the industry was the guy who announced at amusement parks: ‘Please keep your arms and legs inside the vehicle,’” says Matt Silver, a Canadian-born voice actor now in the U.S.
The potential for AI to replace at least some voice work is very real, he says. When it comes to advertising, he expects that agencies will first test the technology to see how audiences respond. If there’s no pushback, some will embrace it. “It’s purely a cost-cutting measure,” he says. “Voice talent is usually one of the last things to be tacked on to a commercial, and one of the most expensive on a per hour rate.”
Silver would consider cloning his voice only in the right circumstances. Recently, he’s had discussions with a company about creating different learning modules for the LGBTQ community, a project requiring multiple tweaks and variations, which an AI voice might handle more efficiently. “It’s a personal choice, at the end of the day, but I do think we’re engaged in a bit of a race to the bottom,” he says.
For David Kaplan, whose throaty baritone has voiced medical equipment, Domino’s Pizza commercials, and at least one Alaskan cruise safety video, artificial intelligence is just another tool, and his AI alter ego will be among those in the Voices.ai library when it debuts later this year. “What that could mean is that I’ll be able to make money while I sleep,” he says. “If somebody wants to hire me right away, and it’s a small-budget job, bingo.”
Kaplan, though, is an established name who has worked in the industry for two decades and faces no shortage of opportunities. He’s recording constantly for clients from his home in New York, often seven days a week. The situation is entirely different for newcomers, and AI will only make it harder for them to break in. “They’re screwed,” Kaplan says. The quick, lower-paying gigs that AI could replace today are often stepping stones for new talent. Why, he asks, hire someone inexperienced when AI David Kaplan is available? “And I don’t even have to show up,” he says.
Younger folks are well aware of this possibility, data from Voices.com show. The company recently surveyed workers in creative professions about their views on generative AI, and found 42 per cent of respondents believe the technology could take over significant elements of their jobs. The youngest creatives, those aged 18 to 34, were the most concerned, whereas older respondents were more sanguine. One-third of those between 35 and 44 had no worry at all.
Tara Strong, who grew up in Toronto and has played characters on animated series such as Rugrats and The Powerpuff Girls, was approached about a year ago by a company that wanted to use an AI version of her voice. Despite the money – “And they offered a lot of money,” she says – Strong turned them down for at least two reasons. “I would be teaching an artificial intelligence to sound like me,” she says, “and, essentially, talking myself out of a job.”
She also worried that her AI self, once brought into existence, could be hijacked without her consent and made to say offensive comments, or simply used in other commercial contexts without her knowledge.
The case of Canadian voice actor Bev Standing is a well-known harbinger for those in the industry. A few years ago, she recorded audio as part of a project to translate Chinese texts. In 2020, she discovered the voice for TikTok’s text-to-speech feature sounded remarkably like her own. Standing sued TikTok’s parent company in 2021, and later reached a settlement.
With the rise of AI, actors and agents are increasingly vigilant about the language in contracts to ensure that a voice cannot be duplicated or used to train an AI model. Unions are catching up, too, although the terms are vague. The Toronto branch of the Alliance of Canadian Cinema, Television and Radio Artists notes in its 2023 operating plan that it will work with international peers to develop an “artist first” approach to AI, and ensure the technology is used ethically.
A looming question is what happens to the value of a human voice in a world of AI clones. Can an actor charge a premium for the real thing? Strong has her doubts. “For the next few years, there’s going to be plenty of work for actors. But down the line, when the technology becomes flawless, maybe voice actors will be obsolete,” she says, adding that an entire ecosystem stands to be affected. If clients can simply choose from a library of AI voices, that could hurt casting agents, too. (Some companies already use artificial intelligence to match actors with clients, bypassing agents.)
But what the most talented voice actors bring to any project is creativity and ingenuity (along with the ability to take direction), none of which AI can do effectively today. That’s the message Kim Hurdon has been reinforcing lately. She runs a casting agency in Toronto and serves as an instructor at a voiceover training school, and recently received a call from a new graduate fretting about the future of the industry.
“My hope is that there will always be the need for humanity to be heard through storytelling,” she says. Projects with smaller budgets are more likely to turn to AI, but actors will still be in demand for premium work, she hopes. “We’re trying to train these voice actors to be at a higher level so they’re competitive,” she says. “When it’s a human with a soul doing a read, it just elevates the work.”