Abby Weffer has been working as a voice actor for three years. She is in the early stages of her career. But she was so prominent that she was anonymized on Twitter in February.
Someone went a step further by creating a fake account that looked just like theirs and put her address on it.
“They were using an AI program to synthesize my voice,” she explains. “And it wasn’t all that bad. … To the layman’s ear, it was pretty close.”
But to make matters worse, “her” voice was blatantly racist. For Weffer, it was like an out-of-body experience. For everyone to hear online, the Evil Twin has been set free on Twitter.
she was worried Will her boss listen? Will she lose her job? To make matters worse, this happened to three of her friends who are also voice actors.
“This kind of defamation is very damaging to people just starting out in their careers as well as to more established people,” she says. “It’s really infuriating.”
In the end, Weffer deleted the tweet (…well, Twitter took three months), but without any career damage she could point out. And although I never knew who posted it, I certainly know how it was made.
There are a number of new AI companies popping up that allow users to upload a few minutes of recorded audio, clone that audio, and have it say something. The imitation isn’t perfect, but it’s safe to say that it quickly becomes indistinguishable from the human voice. Anyone with enough voice is a target. And voice actors produce a lot of voices.
Audio work is everywhere: in comics, commercials, video games, but also audio books and corporate training videos. Then someone has to hail the bus and let them know which stop it is at.
Since ChatGPT came out late last year, we’ve heard a lot about AI and all the good and bad things it might bring us.and i’m going to hear a lot more from AI will make it easier and cheaper for companies to let robots talk.
This raises some important questions, such as who should control this technology. Will the actor itself become obsolete? And who is to blame if it is used against someone?
That’s the problem Susan Eisenberg had to deal with. Eisenberg is a legend in the voice acting world. She has provided the voice of Wonder Woman in the comics ‘Justice League’ and ‘Justice League Her Unlimited’, as well as many video games such as ‘The Elder Scrolls V: Skyrim’.
Like many pop culture titles, this game has a dedicated fan base. And Bethesda Games, the company behind the game, is allowing fans to use that data to create new things (weapons, costumes, scenarios, etc.). These are called mods and are basically fanfiction. They are all hosted on a site called Nexus.
In May, a fan wrote to Eisenberg telling her that some of the characters she played, including her voice, were used to create porn scenes. Results range from embarrassing to disgusting, and while the voice doesn’t speak the way Eisenberg speaks, it definitely sounds like her.
“It got past the upset and I was reluctant to even see what was going on, but I had to because it was my responsibility to see what was being used,” she said. To tell. “It is very unclear to me how to proceed with this matter, because who owns my voice?”
Bethesda Games did not respond to a request for comment. Nexus, the company that hosts all of these mods for him, told me their policy of removing offensive content if requested by the creator or voice actor.
“…if you receive a credible complaint from a party who feels that your work is harming them, this could be the voice actor or the entity that owns the character/asset rights. We will remove the content if necessary.”
In other words, Eisenberg and other actors need to be vigilant, scouring the internet for clones, as it’s up to them to police their likenesses. And for now, legally this is a gray area.
What Weffer and Eisenberg experienced is one of the biggest fears voice actors have about AI. Your voice is duplicated and used against you. Another fear he has is straight out of an old sci-fi story and being completely replaced by machines.
To dig it up, head to the recording studio behind a residential suburb of Melrose. An actor named Matthew Parham is in the booth trying to pull magic out of a human voice box for a commercial.
Whether it’s video games or cartoons or whatever, the company is so secretive that we can’t say exactly what Parham is recording. In this case, we can say that it is a commercial for a product best known as Orange.
Perham began his voice acting career somewhat by accident during the pandemic. He does a lot of commercial work now, he says, “but most… require a ‘cool sound’ or ‘urban’ voiceprint in dog-whistle parlance. It’s a brand.”
Sometimes brands fall for trying not to look racist, but this boils down to the fact that brands aren’t asking someone to “look black”, they’re asking them to look black. Get noticed.
Parham and his director Tim Friedlander (owner of this studio) have a sort of fictional dial of black voices. Rapper Tupac Shakur on the other. On the other is the non-threatening Fresh Prince of Bel-Air character Carlton Banks.
“So you kind of figure out what the brand is and start from there,” he says. “Tech companies’ ‘urban’ productions will seek more of Carlton’s voice than Tupac’s.”
Tim Friedlander conducts Matthew Parham in the recording booth. Photo by Brandon R. Reynolds.
Voice work is an art and a business. Adjust certain attributes, rhythms, accents, atmospheres and adjust others. Successful voice actors are good at instantly tuning in to what their clients want or need, whether it’s the voice of an animated monster, a corporate training video, or a city bus.
But what if a robot could do that? Could someone turn a real dial in the app and add a hint of Tupac in a synthesized voice to summon more Carlton?
It poses an existential threat to the voice actors – just as AI scripts pose to the Writers Guild’s striking writers.
Director Friedlander, who works with Parham at the booth, addresses that threat directly. He is also a voice actor and founder of a non-profit organization called the National Voice Actors Association (NAVA). He started the service as a way to get health insurance for voice actors, but last year he was drawn to the big buzz of artificial intelligence. Specifically, what happens when AI can clone and create voices on demand?
“We basically started thinking about what we could do in that case. What could we do to protect the voice actors?” he says. “What can we do to prevent someone else’s voice from being stolen and used without permission?”
Sofas in Tim Friedlander’s recording studio are lined with decorative pillows, including one with the face of actor Nicolas Cage. Photo by Brandon R. Reynolds.
This is a big part of a potential strike by SAG-Aftra members to keep the studio from going full robot.
“It comes up in almost every discussion we have had about strikes. There is also sex. ’ he says. “So you’re going to put a ton of people out of work…for what?” Saving money seems like that. “
SAG-AFTRA is certainly thinking about this as it continues negotiations with studios and streamers. The current concern is whether the actor owns his image or voice, and whether his company owns it. should do it You get to know that you can clone someone’s voice or image, and how actors get paid for it.
“We need that person’s informed consent as to what we do with their voice models and other data. It’s just their personal data, and it’s obtained as part of the creation of whatever project it is.” It is,” said Duncan Crabtree Ireland, SAG-AFTRA’s National Secretary and Chief Negotiator. “And they need to be fairly compensated for it, because their voice will be the voice of any project going forward.”
Unions do care about signing union actors, but according to a survey conducted by NAVA, most actors are not in a union. (To that end, NAVA has created a covenant that actors can try to insert into their contracts, which prohibits them from using voice data for AI interaction.)
But voice actors are also excited about what AI can do for them. They can use clones of themselves to fill gaps when they can’t record or their voices are blown out. It’s a vocal mini-me that does all the screaming and tiny little edits that no one wants to spend time recording. booth.
Actors then license their voices to use them, for example, to create voices for speech-impaired or non-verbal people, or as voices that automatically convert written text into speech. can be obtained.
Or even do boring things, like Friedlander did earlier in his career.
“I spent six months creating a training manual for GE transportation services on how to repair train diesel engines,” he says. “I spent months reading ‘3/4 turns with a 5/16 wrench, torque values go into this and this and this hose.’ There is one…and if I have a synthesizer or synthetic voice that this company is licensed to use, they can compensate me and pay the license fee for my voice.”
It’s a brave and noisy new world of clones doing a lot of work humans can’t have lung capacity. I asked Perham if he could license my voice.
“I will lend my voice to a synthetic voice for a reasonable price, with the understanding that it will be used with my consent,” he says.
Out of strictly professional interest, I had to know the number.
“It’s enough that I don’t have to work anymore. I don’t know if it’s inflation, but it’s in the millions or so.”
Nice to know, nice to know (write it down).
Voice actors are the canaries in the coal mine for how AI will change our jobs and our online identities. And when they say they’re worried or excited, maybe we should listen.
