Companies insist on using Remie Michelle Clarke’s voice. Award-winning vocalist Her smooth Irish accent has powered ads for Mazda and Mastercard and is the sound of Microsoft’s search engine Bing in Ireland.
But in January, her sound engineer told Michelle Clarke that she had mysteriously found a voice that sounded like her in some unexpected place. For an affordable monthly fee, Revoicer customers have access to hundreds of different voices, voicing commercials, reciting corporate trainings, narrating books, and more via artificial intelligence-powered tools. and transform it to say anything.
Revoicer advertised “Olivia” with a picture of a white-haired woman of presumed Asian descent and the following blurb:wonderful [sic] for audiobooks. ”
38-year-old brunette Michelle Clark looked nothing like “Olivia.” “Hello dear people, my name is Olivia.” “My voice is gentle and caring.”
In an interview with The Washington Post, Michelle Clark said, “It’s just weird. ‘When you see your voice shifted or altered … there’s something very invasive about it.'”
But Michelle Clark isn’t the only one who’s lost control of her voice. advances in generative artificial intelligence, techniques that form text, images, or sounds based on supplied data; It allowed software to recreate people’s voices with eerie accuracy.such software Quickly spot patterns by comparing small samples to a database of millions of voices. It allows users to wield a simple text-to-speech tool to change the voice to say whatever they type.
The technology made headlines this month when a music producer claimed to have used AI versions of Drake and The Weeknd’s voices to create a new track, “Heart on My Sleeve.” This track went viral on his TikTok. A cloned voice of Emma Watson recited Adolf Hitler’s Struggle for Me line, and it was artificially told that President Biden preferred low-grade marijuana.
But the technology puts the unknown voice actors who narrate audiobooks, video games, and commercials in a particularly precarious position. Their voices are well known, but they rarely command the star power needed to control them. Copyright provisions do not address artificial intelligence’s ability to reproduce human-like speech, text, and photographs, so the law offers little refuge. Provisions are often included, allowing companies to use the actors’ voices in endless permutations or even sell them to other parties.
According to Revoicer.com developer Neal Throdes, the company uses voice through a licensing agreement with Microsoft that gives it unlimited access to Michelle Clarke’s samples. Hours after the Post contacted her Revoicer.com, the company promised to remove the voice from the site. “We took responsibility,” Throdes said in an email, adding, “Revoicer.com is not responsible for this situation.” [Michelle Clarke] ____ is inside. ”
Several voice actors told The Post they may abandon their careers, seeing a cataclysmic future where people can get voices without hiring individuals. I wonder why a company would pay $2,000 for a 30-second recording she could order when she could pay $27 a month for a clever clone.
“How many other companies… take advantage of what I say, take advantage of my work, make a living without considering me?” asked Michelle Clark.
‘That’s scary’
Speech-generating software has benefited from the surge in sophistication of generative AI over the past year, powering chatbots such as ChatGPT and text-to-image makers such as DALL-E.
AI has long helped companies mimic voices, but it has created voices that are robotic and unrealistic, said Zohaib Ahmed, CEO of Resemble. AI, a company that uses artificial intelligence to generate speech.
However, improvements in the software’s underlying architecture and computational power have increased its power. We can now quickly analyze millions of sounds to find patterns between basic phonetic units called phonemes. The software compares the original audio sample with similar samples in the library, finds unique characteristics and creates realistic-sounding clones.
Before this advanced pattern recognition was possible, speech-generating software required thousands of sentences to replicate speech, Ahmed said. Today, these tools work with just a few minutes of recorded audio.
“We don’t need an hour, we don’t need 20 more hours,” says Ahmed. “It takes minutes, seconds…basically to get what you hear…90% [accurate]”
This progress has been a boon for some. People with degenerative diseases such as ALS can use artificial intelligence to bank their voices. Val Kilmer, who lost his voice after surgery for laryngeal cancer, was able to use voice-replicating software to to talk about his role in ‘Top Gun: Maverick’.
But it also creates a predatory industry. People report that their loved ones’ voices are recreated to perpetuate the scam. Scavenge the internet for high-quality voice samples, bundle hundreds of voices into a library, and sell to businesses for commercials, in-house training, video game demos, and audiobooks for less than $150/month. Startups are emerging.
Tim Friedlander, president of the National Voice Actors Association, an advocacy group, said these “middleman” startups offer companies a favorable proposition.
Friedlander added that generative AI’s impact on his industry is still in its infancy and is likely to cause major disruption. “It’s scary,” he said. “Voice actors are unknowingly training their successors.”
“That’s my voice”
One afternoon, when Bev Standing was home, her children sent email after email. I ask the same thing: Mom, are you the voice of TikTok?
Standing was confused. The Canadian voice actress has worked for many clients, but TikTok never hired her to voice over, nor did she get paid by her parent company, ByteDance.
But on the app, she found herself everywhere. As the voice behind TikTok’s iconic text-to-speech feature, she’s narrated cat videos, criticized shoddy boyfriends, promoted McDonald’s burgers, and pitched investment tools you’ve never heard of. .
She didn’t immediately get angry. She “had been fun for about three days,” said Standing. “But as soon as my business brain kicked in, it wasn’t.
In 2018, Standing took a job with a China Acoustic Research Institute client to record her voice for a translation app. She read in her monotone style, which is typical of TikTok’s voice-over feature, but she said there was no provision in her contract that allowed her to sell her voice to other companies.
She sued ByteDance in 2021 and settled for an undisclosed amount. Shortly after, TikTok removed her voice from the app. He is voiced by Canadian disc jockey Kat Callahan.
The software that duplicates Standing’s voice is likely not as sophisticated as current technology, but Standing states that he does not welcome having his voice copied without permission.
“That’s my voice,” she said. “You cannot receive it without paying me.”
Despite Revoicer.com promising to remove “Olivia’s” voice, Michelle Clark says her livelihood is still in jeopardy. Other third party her sites may be reselling her voice. Her friend passed her an Instagram ad that she appeared to be narrating, even though she had never heard of the company: “The problem is unsolved for me,” she said. said.
However, as the mother of a 1-year-old boy, she thinks she may quit voice-over work. “There’s no right time to feel like your future is at stake,” she said. “But now is the absolute worst time for me.”
We rarely rely on voice actors. Until recently, artificial intelligence did not pose a major threat to their profession, and many dissected contracts looking for provisions that would allow companies to use audio beyond their individual jobs. was not
Copyright law is also immature, does not determine what happens when a person’s voice is imitated for profit, and celebrities have access to more protection than lesser-known professionals. (Drake’s AI-generated song, for example, was quickly removed from YouTube and Spotify last week after Universal Music Group raised concerns.)
Daniel J. Gervais, an intellectual property expert and professor at the Vanderbilt University Law School, says U.S. law doesn’t provide enough refuge for those who have been deprived of their voices. .
Federal copyright law does not protect individual voices, and local laws vary from state to state, he said. even, it is difficult to claim who is covered. says Gervais.
Friedlander said colleagues should be vigilant about how their voices are being used on the internet and pay close attention to contract details.
Many voice actors are not unionized, and Friedlander’s advocacy groups are urging voice actors to look for clauses that call for permanent rights to their voices. We created a template contract to give you control over how your voice is used.
In Europe, it’s easier to copyright recordings, and commercial scraping of such content requires permission from the recording owner, Gervais said.The European Union also takes a stronger stance on artificial intelligence by proposing legislation Classify the risks of AI systems.
“There is a big fork in the road between Europe and America,” he said. “It’s much more aggressive.”
“There’s nothing better than me”
In late January, Mike Cooper received an email from a company advertising a library of voice-overs for sale. Intrigued, he scrolled through the pages and soon found his own voice in the library as a sample.
“So when I clicked ‘play’, it was a very surreal moment to hear my voice come back,” he said.
Cooper, who lives in Asheville, North Carolina, said he was angry at first. However, he remembered why this happened. The company that sells his voice now, he probably got it after Cooper bought the company that did his few minutes of voice-over work in 2016.
Cooper remembers a clause in his contract that his voice could be used elsewhere. He only gave the company his voice for a few minutes, he said.
“We thought the risk was very small,” he said. “I was completely wrong.”
But Cooper says an artificially generated voice created without his own input can’t provide a deep understanding of what the project needs and a performance with emotion and intent. .
“It’s not as good as I am,” he said.
