“The effect is, frankly, catastrophic” – the voice actors and unions fighting back against gaming’s controversial use of AI
Can there really be an ethical use of AI when it comes to actors in video games? Developers are increasingly experimenting with the technology, but what does that mean for the profession of voice acting?
Lobbying against the practice was certainly a key part of the SAG-AFTRA strike that finally concluded in the summer, with the new contract ensuring “safety guardrails” around AI technology. Yet while this agreement is meant to protect the livelihoods of actors, AI remains in use to voice some video game characters. Just earlier this month, Let it Die: Inferno developer Supertrick Games admitted generative AI had been used for two character voices as a deliberate, creative choice. It claimed these voices “are not derived from or modeled after any human performers, ensuring no copyright concerns”, though presumably these voices still came from somewhere.
The most high profile recent example was Arc Raiders, criticised in Eurogamer’s review for its use of AI voices. Embark Studios has been open about its use of AI in game development, and its previous game The Finals also utilised AI voices. For Arc Raiders, human actors trained an AI text-to-speech tool utilised for incidental NPC dialogue (but not key story points). Those actors reportedly gave consent and were paid accordingly. But despite its ancillary usage, does this set a worrying precedent the rest of the industry could follow?
Manage cookie settings
ARC Raiders ReviewWatch on YouTube
I wanted to understand the varying perspectives on this controversial practice, to understand more about the financial and ethical concerns of AI usage. As such, I’ve spoken to unions, studios, and – of course – actors on how AI is impacting the future of voice acting.
But first, why did Embark Studios choose to use AI in the first place? The studio did not respond to Eurogamer’s request for comment, but ahead of the release of Arc Raiders CCO Stefan Strandberg previously explained to Eurogamer how the studio intends to be on the “forefront of emerging technologies”.
On text-to-speech, Strandberg discussed the use of real actors. “There’s something special and dynamic about putting two people together that to me, being a sound designer originally, there’s nothing that can replace that in my opinion,” he said. “TTS allows us to increase the scope of the game in some areas where we think it’s needed, or where there’s tedious repetition, in situations where the voice actors may not see it as valuable work. So it’s a wide umbrella, but the experience of the game doesn’t use any generative AI.”
Embark used AI for some NPC voicelines in Arc Raiders. | Image credit: Eurogamer / Embark
The use of such technology has helped Embark Studios remain a lean team, and allowed them to quickly add or alter voice lines without the need to re-record. But as Strandberg surmised: “There are no shortcuts to making great games. You need so many things in place: a clear vision, a great culture, clear goals. The vision will inform how you get there.”
Still, text-to-speech is a form of AI that may be based on human voices, but the final result is, arguably, not human, devoid of the nuance a human actor can provide through proper performance.
One general assumption is that the use of this sort of AI technology is cheaper than hiring voice actors for full voice work, allowing studios to save money. Yet voice actors are already severely underpaid for their work, and US actors’ union SAG-AFTRA is fighting not only for better contracts but to ensure actors are equally remunerated when AI is involved.
The union’s current standard rates, published publicly, are $1,135 for an off-camera day performer of up to three voices, or $567.50 for one voice. Additional fees can be applied, including if a role requires full motion capture, plus some actors can of course negotiate higher fees – this is just a union recommended minimum.
Speaking to Eurogamer, the union’s national executive director and chief negotiator Duncan Crabtree-Ireland confirms SAG-AFTRA’s agreement requires actors be paid on an equivalent basis for digitally replicated voices. What’s more, the union specifically bargained over real-time generation of voices using digital replicas, resulting in a minimum negotiable rate of 750 percent of scale (that would be $8,512.50 based on the $1,135 fee). As such, AI voices are perhaps more likely to be implemented for creative reasons, or speed of development, rather than financial concerns.
“None of this can or should be done without the consent of the performer, and we’ve established that firmly in our collective bargaining agreements, and frankly, outside of our bargaining agreements,” says Crabtree-Ireland. “We’ve been working hard to get that principle established in the law, so that even in non-union settings, that minimum standard is present.”
“None of this can or should be done without the consent of the performer.”
He continues: “Our long term goal is to make sure that there’s a clear respect for people’s right to control the use of their image, likeness and voice for purposes of any kind of replication, especially AI. And I think we will ultimately achieve that, notwithstanding the AI regulation skittishness that some governments have, because we’re not talking about regulating AI. What we’re talking about is giving people the right to control the use of their image, likeness and voice.”
Here in the UK, actors’ union Equity also has recommended rates, though these are published only to members. “It’s usually the case that games companies don’t engage performers directly, but contract studios who own recording spaces and equipment, and who in turn engage performers,” Equity audio official Shannon Sailing explains to Eurogamer.
“As there are no union agreements with games companies (via studios) in the UK to set minimum standards for performance work in games, the rates vary across the board. This is wrong. It’s the workers – including performers – who are at the heart of creating a game, and they deserve proper pay, terms and conditions set by a union agreement.”
The use of AI “leads to a subpar experience for gamers,” Sailing adds. “No machine can capture the authenticity or emotion of a human voice actor.” Still, the union understands the use of this technology won’t disappear anytime soon. As such, it urges for consent, transparency, and proper remuneration for actors.
“No machine can capture the authenticity or emotion of a human voice actor.”
What, then, of recording studios? Is AI impacting their work?
“I think it would be very naïve to say that AI voice isn’t going to have an impact on the amount of work available to actors, but studios big and small are faced with the same dilemma every year: do more with less,” says Chris Donnelly, director of audio and datasets at Side UK – the studio behind the award-winning voicework in the likes of Clair Obscur: Expedition 33, Silent Hill 2, Still Wakes The Deep, Final Fantasy 14, Baldur’s Gate 3, and many more.
“As creatives, we have to ensure that the call for ethical practices, strong guardrails and robust legislation is heard, as well as the financial arguments. That means fair remuneration for the actors and full transparency of how the model will be used, delivering clear and practical benefits for game developers that result in high-quality performances that gamers rightfully expect.”
Side is behind the voice work in many of your favourite games. | Image credit: Larian.
Side is currently setting up a global committee to support client requests for AI content across its business, Donnelly explains, and the studio already has a “human in the loop” policy for any AI-related requests. “Our overall hope is that using these technologies in an ethical way, will mean that opportunities expand, adding additional depth and immersiveness to games on top of what might have been done historically,” Donnelly says.
He admits the use of AI is “inevitable” to keep up with the demands of players for immersive experiences with thousands of realistic, dynamic NPCs, something that simply isn’t possible to record using traditional methods. “Having said that,” says Donnelly, “technological progress is complicated and for Side as an audio studio, we know that this is going to change the way we work. What I do feel absolutely certain of, is that AI voices or not, artists, actors and storytellers will all still be needed to make games happen.”
While unions and studios are fighting for regulation and ethical practices, it’s actors themselves who must suffer the human cost of increased AI voice usage. Ultimately, perhaps, it’s down to them to choose whether or not to accept contracts with AI clauses included, though if the practice grows will they even have a choice?
“I see why devs and publishers are tempted by new models that let them change the scope of the narrative with a few keystrokes,” says David Menkin (Final Fantasy 16, Lego Star Wars: The Skywalker Saga). “I also get why actors are tempted by a paycheck. But when creative labour becomes data, a performance can easily stop being something interpreted and start being automated.”
Menkin says AI must be used ethically to support and not exploit creatives. “Contracts are rarely written in favour of those at the coalface and it won’t take much for this new type of tech to be abused,” he says. Regulation is needed, but despite performers pushing back via unions, “nothing will effectively change without government or at least industry involvement.”
“When creative labour becomes data, a performance can easily stop being something interpreted and start being automated.”
Jennifer Hale (Mass Effect, Metal Gear Solid) similarly calls for consent, control, and compensation, adding actors “should be compensated for what would’ve otherwise been paid work for them, especially when that work is used to generate profits for someone else.”
Hale also highlights newer actors, without the knowledge to negotiate, will make “fear-based decisions and underprice themselves”, which is why union status and open communication about fair rates is so important. “It’s more imperative than ever that we have each other’s backs and that we’re clear about where we bring value and what we receive in exchange for that value,” she says, adding: “I believe having human actors may well become an important mark of quality that elevates some games above others.”
What’s more, actors can “choose not to ‘other’ this technology,” says Hale, but “move forward with it in a conscious, collaborative, constructive way”. Alix Wilton Regan (Tomb Raider, Cyberpunk, Dragon Age) is more staunchly against AI technology.
“The effect is, frankly, catastrophic,” she says of AI’s impact, noting “the voice acting market is drying up and wages are stagnating” as the tech can result in a reduced fee. “Already, many talented performers – of all ages – can barely pay their bills,” she says. “The supply of actors is growing, but the demand for actors is shrinking, as corporations look to replace us with AI.”
Wilton Regan will voice Lara Croft in the next Tomb Raider game | Image credit: Crystal Dynamics
For Wilton Regan, there are no benefits to the use of AI, even for more extreme voicework like incidental grunts and yells. “I love being in the studio and collaborating with different teams as we build new worlds and characters together, so, at best, AI alleviates some minor inconveniences such as the potential for vocal stress,” she says. “But it’s at the cost of endangering our livelihoods. That’s a horrible trade-off that no one should have to make.”
Colin Ryan (Final Fantasy 14, Hades 2) believes similarly, but says he would “prefer to see investment in vocal support for actors in games”. Where theatre, TV and film have voice coaches to support actors with challenging moments, “that’s not an option in video games.”
“Developers using AI as a shortcut and to further bring down costs is a massive blow to actors’ livelihoods and craft,” Ryan continues, adding Embark Studios using AI even for incidental NPCs sets a “worrying precedent”. “For many actors starting out, getting to record NPCs is a way to break into the industry, an opportunity for them to hone their craft, and form career-spanning connections, just like it was for me,” he says. “To take away these jobs and use AI to generate them shows how little they value our contribution.”
Further, Ryan believes the use of AI voices “isn’t a creative decision, it’s a financial one”, and while he admits it often sounds hyperbolic and alarmist, “the truth is there will be less and less commercial incentive to hire humans” as the technology improves. “As an industry we need to make the case for humans and human creativity, to ensure that actors are a valued and intrinsic part of video game creation,” he says.
Ryan plays Apollo in Supergiant’s latest hit | Image credit: Eurogamer / Supergiant
However, Wilton Regan is fighting back against AI “abuse”. Along with author Nolan Kelly, she’s launched TIVA (True Indie Voice Art) as an audiobook project using poison-pilling technology to ensure the work of actors is unusable for AI. Its cast includes the likes of Hale, Ashly Burch (Horizon), Jane Perry (Returnal), and Devora Wilde and Jennifer English (Baldur’s Gate 3), among others. “Our hope is that the quality of this work will convince people of our value as actors and the value of our profession in the voiceover space,” says Wilton Regan.
The project is just one example of how actors are standing up for themselves against the threat of AI, but it’s clear wider government regulation on AI technology is required in addition to union efforts.
“As an industry we need to make the case for humans and human creativity, to ensure that actors are a valued and intrinsic part of video game creation.”
Whether for perceived financial reasons, development shortcuts, or as a creative decision, video game studios are increasingly looking to AI for voicework. And as awards-season hits and we all seek to celebrate the work of video game performers, it’s imperative the industry remembers to protect the creative work of actors. Otherwise, it’s a slippery slope towards a huge drop in quality.
“Even if human performers provide the base voices, and even if the makers have the best of intentions, without an actor bringing a character to life, you simply have a creepy cheap human imitation performing at you,” says Wilton Regan. “Not for you, but at you. For anything generated, there is no true emotion, no chance for repeat takes, no nuance that a talented actor can bring to the role. You need a human for that.”
