‘It’s very easy to steal someone’s voice’: how AI is affecting video game actors
The increased use of AI to replicate the voice and movements of actors has benefits but some are concerned over how and when it might be used and who might be left short-changed
www.theguardian.com
‘A lot of people can’t even envision the extent of disruption there will be in the near future.’ Photograph: Dino Fracchia/Alamy
Artificial intelligence (AI)
‘It’s very easy to steal someone’s voice’: how AI is affecting video game actors
The increased use of AI to replicate the voice and movements of actors has benefits but some are concerned over how and when it might be used and who might be left short-changedDavid Smith in Washington
Fri 29 Mar 2024 06.02 EDT
Share
When she discovered her voice had been uploaded to multiple websites without her consent, the actor Cissy Jones told them to take it down immediately. Some complied. “Others who have more money in their banks basically sent me the email equivalent of a digital middle finger and said: don’t care,” Jones recalls by phone.
“That was the genesis for me to start talking to friends of mine about: listen, how do we do this the right way? How do we understand that the genie is out of the bottle and find a way to be a part of the conversation or we will get systematically annihilated? I know that sounds dramatic but, given how easy it is to steal a person’s voice, it’s not far off the mark.”
https://www.theguardian.com/games/2024/mar/23/game-developer-conference-layoffs-ai-support
Jones, 45, a voice artist with credits including Starfield and Baldur’s Gate III, was wrestling with the march of artificial intelligence (AI) into video games, increasingly recognised as less a niche pursuit for bedroom-dwelling teenagers than a storytelling platform with almost unlimited potential. Hollywood actors such as Jodie Comer, Idris Elba, Megan Fox, David Harbour and Keri Russell are contributing their likenesses and voices to the multibillion-dollar industry.
Just as in film and TV, only more so, AI represents a gathering storm for video game actors. Some studios are experimenting with tools that can clone voices, alter voices and generate audio from text. In interactive, multi-choice games, this can generate a potentially endless number of characters and conversations – and is far more efficient than asking performers to record huge quantities of dialogue.
The response from professional actors has been mixed. Some fear that games companies – sensing opportunity to cut costs and accelerate development – would use AI to reproduce their voices without permission or payment, pushing down the value of their work. Others have been willing to give it a try if they are fairly compensated and their voices are not misused.
Jones, for her part, had a brainstorming session with colleagues for a few months and came up with a structure for an AI company that could coexist with actors. She is now co-founder and vice-president of strategic partnerships at Morpheme, a startup aiming to harness AI to reshape how vocal performances are used in everything from animated series to video games.
Morpheme’s AI software records audio from actors and then creates a model of their voice that can be used to alter, expand and enliven future productions. It has been demonstrating the technology to several top gaming companies.
Gamers play video games during the Gamescom LAN event. Photograph: Ina Fassbender/AFP/Getty Images
“We’ve been going full steam ahead, creating contracts that work for actors, making sure that actors understand if they want to record with us, if they want to have a digital double, number one, we get their consent. You want to have a digital version of your voice? Fantastic. We pay them and then any time the voice is generated they also receive payment. In addition, if at any point they no longer feel comfortable having their voice be a part of our offering, we will delete it.”
Unlike their counterparts in film or TV, voice actors for video games do not receive residual payments after their recording sessions. Some gaming actors are looking at the emerging AI technology as an opportunity to potentially collect extra payments down the road on top of a base minimum. Under Morpheme’s contract, actors who are unavailable or unable to work on a new project can put their “digital twin” to work, and, in exchange, receive additional money.
But not everyone is ready and willing to play by the same rules. Jones was recently offered a job for a one-off fee but then found, buried in an 11-page contract, an option for the employer to create a digital version of her voice for use in perpetuity without any additional payment. Unauthorised uses of AI technology are already proliferating, as illustrated by a recent hoax Joe Biden robocall and deepfake recordings of the actor Emma Watson reading Adolf Hitler’s Mein Kampf.
Jones, who is based in Los Angeles and has worked on about 300 games, notes: “It is very easy to steal a person’s voice. At the beginning of 2022 it took six hours. At the beginning of 2023 it took three hours. Do you want to guess what it takes right now? Three seconds. Anything you have on Instagram, TikTok, any YouTube videos, anybody can create a digital version of your voice from just that. Is it perfect? No but the technology is not getting worse.”
She adds: “The danger is that people can take all of these billions of voices that are available online, scrape the internet for them, mush them together and create a new voice that does not ‘belong’ to anybody, thereby creating a ‘new’ voice. However, they are still profiting off of my voice.
“We’re working on active fingerprinting technology that could parse that out but, as quickly as we’re working on developing that companies are working to erase that. It’s the old network security versus hacker problem. As soon as network security figures out a lock, hackers figure out a way through it.”
Jones also sits on the board of the National Association of Voice Actors (Nava), a non-profit which has a mantra of “consent, compensation and control” around the use of AI and has been in talks with with members of Congress on upcoming AI legislation. “We’ve been working with the Office of Copyright because right now you can copyright your name, image and likeness – you cannot copyright your voice.”
There are concerns that AI voices could replace all but the most famous human actors and eliminate entire job categories, such as quality-assurance testers or the entry-level positions that allow young performers to get a foot in the door. Some actors worry that they might already have signed their voice away years ago and have no way of claiming it back.
Tim Friedlander, an award-winning voice actor who is founder and president of Nava, says: “There is fear. There is uncertainty. There is kind of a helplessness: how do we, as independent voice actors who are in the union or not in the union, push back against multibillion-dollar companies who have the ability to outspend us and out-lawyer us and potentially – through predatory behaviour or predatory contracts – take advantage of voice actors?
“If you’re under a union contract, you still have to read your contracts, make sure that there’s no addendum or added language that is in there. As voice actors we’re not lawyers, we’re not contract specialists. It is potentially the fear of many people that they’ve given away their voices years ago through contracts, that the damage has been done already and we’re just now going to start to see the results of those predatory contracts from years ago.”