AI in Voiceover: Where Games, Dubbing, Ads and Animation Are Feeling the Impact
AI is changing voice work, but not every part of the industry is changing in the same way or at the same speed.
In Episode 4 of HIGH NOTES, Melissa Thom speaks with London-based actor and voiceover artist David Menkin, whose work spans film, television, commercials, dubbing, animation and games, including Lego Star Wars, Final Fantasy XVI and Pragmata. With more than 20 years as a performer, David has seen technological shifts from several sides of the industry and has become increasingly involved in advocacy around AI, voice rights, and performer protection.
The conversation moves sector by sector, looking at where synthetic voice is already replacing human work, why some areas are more vulnerable than others, and what performers can do when the law and contracts have not yet caught up with the technology.
Prefer to listen or watch? Hear Melissa’s full conversation with David Menkin in Episode 4 of HIGH NOTES, where they explore how AI risk differs across voiceover sectors and what performers can do to protect their work.
AI is not affecting every type of voice work equally
Asked which areas have been hit hardest, David points first to corporate and medical voiceover.
He describes these areas as particularly vulnerable because they often involve measured, informative speech. Synthetic voices have also been used as placeholders and timing tools for years, so producers and clients have become accustomed to hearing them during production. Once the quality is acceptable, the commercial argument for replacing a human voice can be fairly straightforward, especially when the audience has already chosen to watch or listen to the content.
David gives a simple example of the pressure behind that decision: if only three words need changing, does a producer really want to bring an actor back into the studio for another session?
For him, practical considerations such as budget, speed, and convenience start to shape creative decisions.
That pressure is now moving into commercial work too.
David says shorter pieces of content are particularly easy to replace because the amount of synthetic speech that needs to be generated, edited and checked is relatively small. A 30-second advertisement presents a very different technical challenge from a game containing perhaps hundreds of hours of dialogue.
He also says performers are feeling that pressure economically. Fewer campaigns, lower rates, and clients using AI as an alternative in negotiations are driving this.
At the same time, replacing a human voice doesn't necessarily eliminate all the work. David points out that synthetic speech can still introduce errors, such as mispronouncing local place names, which then creates a need for people who can identify and correct those mistakes. The saving is therefore not always as simple as removing the performer.
Why dubbing and localisation face a different kind of pressure
Dubbing presents another set of circumstances.
For David, the driving force here is volume. Streaming platforms and social media have created an enormous amount of content that needs to travel between languages and territories. Dubbing has always involved one voice standing in for another, so the conceptual step toward an AI-generated replacement is relatively small.
But that doesn't mean every market is approaching the problem the same way.
David points to Italy, where he says dubbing performers recognised the risk early. He describes contracts used by major streamers and distributors that prevent dubbing performances from training AI, along with transparency measures intended to make it harder to mislead audiences about synthetic content.
His interest isn't simply whether AI can perform the task. It is in the rules surrounding how the original human performance can be used, and whether performers know and agree to what is happening to their work.
Games present both opportunity and risk
Games create another very different problem because of their scale.
David explains that only a few years ago, creating a stable synthetic model could require a very large amount of recorded material. Now, in his view, far less material may be needed.
At the same time, games are becoming larger and more open-ended. Developers want characters that can respond dynamically to players, create new dialogue, and potentially keep generating material long after the actor finishes recording.
David sees a legitimate commercial opportunity here. A performer could choose to license a voice or performance for this kind of use.
The problem, he argues, is that the legal and contractual protections surrounding that opportunity remain inadequate. The technology may allow a performance to be extended, modified or regenerated, but performers still need to understand who controls that model, how it can be used and what happens to it afterwards.
That distinction runs throughout the conversation. David is not arguing that synthetic voice can never be used. The issue is whether the person whose work makes that synthetic performance possible has meaningful control over it.
When your voice starts saying things you never said
For performers, the risks can become much more personal than losing a job.
David describes colleagues discovering that recordings from games had been gathered by AI aggregation sites and then used to create material they had never performed, including racist, violent and sexualised content.
He has experienced this himself. His nephew sent him recordings of what David describes as a facsimile of his voice saying racist and misogynistic things. Although the imitation was crude, he could hear that his existing work had apparently been used to create it.
For David, that shifts the discussion from copyright or income to identity and harm.
He argues that traditional UK copyright law was created to address a very different problem: copying existing films, music, television programmes, and other works. It was not designed for a world where technology could learn to imitate how someone looks, moves, or sounds, then generate something that person never actually did.
That is why David and other performers and campaigners have been pushing for what he calls a “transformative right”: clearer control over the use of a person’s voice, face, and likeness when those things are used to create something new.
As David puts it, the principle is that this is your voice and your face, and someone shouldn't be able to use them to build something else without your consent, especially through rights hidden in a broad buyout clause.
Would you know what to look for in an AI casting call?
One of the most practical parts of the conversation comes from a real AI-related games casting call David had recently reviewed.
On the surface, he says, the licensing language looked almost reasonable. But when he and colleagues examined it line by line, they found questions that an actor unfamiliar with AI contracts might not realise they needed to ask.
Who owns the voice model once it's created?
How long does the licence last?
Can consent be withdrawn?
What happens if the game generates dialogue that the actor never recorded?
How much control does the performer retain?
And who is the real end client?
David is careful not to suggest the casting director or developer deliberately tried to exploit performers. His concern is that when regulation is limited, the individual actor bears much of the responsibility to understand the agreement before signing it.
That is why he believes actors need to ask questions before they accept the work, not after.
Due diligence is becoming part of the job
That can be uncomfortable for performers who entered the industry to perform rather than negotiate contracts.
Melissa and David both acknowledge that many actors do not enjoy negotiation and can feel overwhelmed by the amount of new information they are expected to understand.
David has experienced that learning process alongside his own agents. He recalls one of them acknowledging that David had begun to know more about this particular area than they did, and choosing to learn rather than ignore the problem.
David now scrutinises contracts far more closely. He describes checking the wording, questioning duplicated clauses, and asking his agent to go back to companies when something doesn't make sense.
His distinction between being a consumer and being a performer is useful. Like most people, he may scroll through the terms and conditions when installing software and simply click agree. But when his own voice, performance, or likeness is the product being supplied, the consequences can be very different.
For performers, understanding what they are granting has increasingly become part of running the business.
Voice actors may simply be experiencing the problem first
One of the wider themes running through Season 4 is that AI voice is not ultimately only a voiceover-industry issue.
Melissa describes voice actors as possible “canaries in the coal mine”: people whose voices have already been recorded and distributed in enormous quantities, meaning they have encountered some of these problems before the wider public.
David agrees.
He recounts talking to an MP about the possibility of AI-enabled impersonation scams and suggesting that families consider agreeing on a private password. If someone receives an urgent call apparently from a family member asking for money, the password offers another way of checking that the caller really is who they claim to be.
For David, examples like this change the conversation. Suddenly the issue is no longer about whether a professional voice actor loses a commercial. It is about what can happen when anyone’s voice becomes a tool for impersonation.
Once people recognise that possibility, he says, the reaction becomes: this affects me too.
This is also why he believes the debate needs to move beyond arguments about performers’ earnings.
David says the more important question is harm: what happens when someone’s voice, likeness, or work can be reproduced and used in ways they did not intend?
He also acknowledges the personal cost of campaigning on the subject. Trying to understand the technology, communicate the risks, and keep working as a performer became almost another job.
But he also sees a positive change. More performers are talking to each other, sharing information and treating the issue collectively rather than trying to solve everything individually.
What happens to the next generation of performers?
Despite the concerns raised throughout the conversation, David remains hopeful about human voice performance.
He believes people still want human connection and that audiences and producers will continue to want a human performer, even if technology becomes appropriate for other kinds of work.
His concern is what happens to the jobs through which new performers traditionally learn their craft.
David describes the lower-paid, less glamorous jobs he did earlier in his career as a kind of apprenticeship. They may not have been prestigious, but they gave him hours of experience in studios, behind microphones and in front of audiences.
If those entry-level jobs disappear first because they are easiest to automate, the industry may eventually face a different problem: where do future experienced performers come from?
Melissa makes the same connection with her early broadcasting work. Craft develops over time, often through work that may not seem important when you are doing it.
Consent, control and compensation
David finishes with a framework that brings many of these questions together: the three Cs.
Consent. Control. Compensation.
Know what you are consenting to when your voice, image or performance is used.
Maintain control with clear licensing terms, including whether and how you can withdraw consent.
And if a model built from your work continues creating value, make sure the agreement addresses how you will be compensated.
For David, AI and human performance do not have to be mutually exclusive. What matters is whether the performer remains part of the decision.
That may ultimately be the thread connecting games, dubbing, commercials, animation and every other area discussed in this episode. The technology and economics differ from sector to sector, but the underlying questions remain remarkably similar: who knows what is happening, who gets to decide, and who benefits from what is created?
Listen to Episode 4 of HIGH NOTES, Across All Sectors: AI in Games, Dubbing, Ads and Animation, with voice and screen actor David Menkin, for the full conversation.
Comments