The loss did not arrive with a famous character replaced on a big screen. It arrived in the ordinary bookings that paid rent between the famous jobs. In October 2026, Simon Kennedy, president of the Australian Association of Voice Actors, told a parliamentary hearing on AI that members had reported a 62 percent decline in available work over twelve months, with synthetic voices taking work once done by humans. At the same hearing, actor Jacqui Duncan said her work had fallen a lot over about eighteen months and that she had taken another job.
That figure is an association report, not an audited national count, and Kennedy presented it as what members reported. Still, it matches a pattern now recorded in three places at once. A large survey in the United States found actors who know a synthetic voice took their booking. A court filing in New York put that loss on the legal record. New union contracts in Los Angeles now spend pages on consent for a replica, because the replica is no longer a thought experiment.
The Bigger Story
The National Association of Voice Actors, known as NAVA, ran its State of Voiceover in 2026 survey from January 22 to February 22. It collected 1,379 responses, spread through social media, voice actor groups, and the NAVA contact list. Asked if they had ever knowingly lost a job to a synthetic voice, 21 percent said yes.
One in five is not a collapse. It is something more useful than a panic number. It is a floor count of loss that actors can name, a booking they held or chased and then watched go to software. The true share will include jobs that simply never arrived, castings that never went out, and clients who stopped calling. No survey can count silence.
On October 29, 2026, that number entered a court fight about something else entirely. SAG-AFTRA, the American Federation of Musicians, and both branches of the Writers Guild filed a motion to submit a friend of the court brief in the consolidated OpenAI copyright case in New York. Page 19 ties synthetic voices to lost bookings. It also cites reporting on short dramas made for phone apps, where actors said they were booked for jobs and then learned AI had replaced them.
The brief matters because of where it puts the harm. The case is about training and fair use, the rule that lets some copying happen without payment. The unions argue the harm falls not just on big news firms and famous authors, but on the people whose voices and performances sit inside the training data. A ruling that training is always fair use, they warn, would wipe out deals they have already struck at the bargaining table.
Context
Voice work was always split in two, and only one half was glamorous. The top half is characters people can name, in animation, games, and film dubbing. The bottom half is narration for training courses, explainer videos, phone systems, adverts for local firms, and rough work that lets a production hear the words aloud. That bottom half paid steadily, and it gave new actors the hours they needed to get good.
That is the half synthetic voice does best. A synthetic voice, meaning speech made by software rather than recorded from a person in a session, is cheap, fast, and always available. It does not need a studio, a director, or a second take booked next week. For a client who only needs words read clearly, good enough now beats excellent next Tuesday. Actors in North Carolina told a student paper the same story this autumn, that tools had taken many low budget and early draft gigs, the jobs that once helped new performers build a showreel and gain skill.
The famous cases sit on top of that quiet shift and make it visible. In May 2025, Fortnite added a Darth Vader that players could talk with, voiced by AI built to sound like James Earl Jones. On May 19, 2025, SAG-AFTRA filed an unfair labor practice charge against Llama Productions, the Epic Games firm behind Fortnite. The union said it welcomed estates controlling digital replicas, but objected that the firm had chosen to replace the work of human performers with AI without notice and without bargaining. Its point was narrow and telling. Other actors had done the work of matching Vader in games before. That work was bargaining unit work, and it vanished without a conversation.
Europe shows the dubbing version of the same fight. Reporting by Reuters in July 2025 followed Boris Rehlinger, the French voice of Ben Affleck, Joaquin Phoenix, and Puss in Boots, and a member of the French effort called TouchePasMaVF, which defends human made dubbing. Dubbing is a whole craft in France and Germany, with actors, translators, adapters, directors, and engineers shaping a performance audiences barely notice as foreign. In Germany, twelve dubbing actors went viral on TikTok in March 2025 with a plea to protect artistic rather than artificial intelligence, and a petition from the VDS voice actors association drew more than 75,500 signatures calling for consent, fair pay, and clear labels on AI content.
The famous voice is the headline. The steady voice was the livelihood.
Different Perspectives
Rehlinger gave the most honest line in the European reporting, and it cuts against easy drama.
“I feel threatened even though my voice hasn’t been replaced by AI yet.”
Boris Rehlinger, speaking to Reuters, July 2025
That sentence holds both truths at once. Many named actors have not been replaced. Many unnamed actors have lost the work that kept them in the trade. Both can be true because the pressure lands at the bottom first and climbs slowly.
The licensing case deserves a fair hearing too. NAVA found that only 13 percent of respondents had willingly let a third party make a synthetic version of their voice, and 94 percent had not made a synthetic version at all. Yet 53 percent said they would be open to one made in an ethical way in future. That is not a workforce refusing a tool on principle. That is a workforce asking for consent, control, and pay before the tool uses them. Voice firms make a matching claim, that licensed clones can pay actors when their voice is used and can open work in languages or volumes one person could never record alone. Where consent and payment are real and tracked, that claim has force.
The contracts show what that bargain looks like when it is written down. SAG-AFTRA members ratified new television animation deals on October 29, 2026, running from July 1, 2026 to June 30, 2030, with pay rises totalling 12.6 percent. A producer now needs specific consent before using a digital replica to dub a performance into another language, a reversal of the old rule. A replica made outside the production still earns at least a session fee, plus residuals where a human performance would earn them. For a synthetic voice that copies no one in particular, producers must notify the union and show the synthetic brings significant added value over a human or a replica. The deal states a principle in plain terms, strongly favoring human performances. A year earlier, the union ended an eleven month video game strike with a 2025 deal built on consent and disclosure for replicas, and the right to suspend consent for new material during a strike.
The limits are just as plain. Those deals cover union work in the United States. Most voice work is not union work. It is booked direct, recorded at home, and owned by the client as a finished file. No contract clause in Los Angeles reaches a training video commissioned in another country from a freelancer found online. The Australian and European fights are running on different laws, different unions, and in the EU case, a plea for new rules rather than a signed deal.
There is also a quality argument, and it should not be waved away. A directed human performance can shift meaning with a pause. It can play a joke, hide a feeling, or land grief without stating it. Synthetic speech has grown far better at clarity and pace, and far less reliable at that kind of choice. That gap protects character work for now. It protects narration far less, because narration was often bought for clarity alone.
Why It Matters
A trade can lose its floor and keep its ceiling, for a while. The stars of animation and dubbing still work, still sign deals, still lend their names to campaigns. Below them, the entry route narrows. If training videos, phone prompts, and early draft reads no longer hire beginners, fewer beginners survive long enough to become the next generation of character actors. The loss compounds in private, one closed home studio at a time.
The consent question reaches beyond acting, into a pattern we have tracked with synthetic companions in Your AI Girlfriend Has a Changelog and with dead performers who keep posting in The Dead Are Posting Now. A voice is not just data about a person. For a voice actor, it is the person at work. When a recording made for one client can train a system that books the next hundred jobs, the old idea that payment ends at delivery stops making sense. That is why the unions fought to move the fight from price to permission.
Audiences will feel the shift in small ways before they name it. More narration will sound smooth, neutral, and hard to place. More dubbed shows will arrive fast in more languages, some of them well made with humans guiding the tools, some of them flat in a way viewers sense but cannot explain. The risk is not that every voice becomes fake. The risk is that real voices become a premium extra, booked only where a producer thinks listeners will notice, much as live music became an event rather than a default.
Closing Reflection
The saddest detail in the NAVA numbers is not the 21 percent who lost a job. It is that they knew. A synthetic voice took a booking and the actor found out, which means the work still existed, the words still needed saying, and someone still got paid to say them. The pay just went to the owner of the software instead of the owner of the voice.
Voice actors are asking for a modest thing that sounds radical only because the technology moved first. Ask before you copy me. Pay me when the copy works. Let me say no and have that no hold after I die, after I strike, and after my session ends. The new contracts prove those terms can be written, priced, and signed. The open question is whether the vast market outside those contracts treats that proof as a standard to follow, or as a fence around a shrinking island, while the water keeps rising everywhere else.
Reporting draws on NAVA’s State of Voiceover in 2026 survey, the unions’ proposed brief filed October 29, 2026 in the OpenAI copyright litigation, SAG-AFTRA statements and contract summaries from 2025 and 2026, Reuters reporting on European dubbing from July 2025, and reporting on the Australian parliamentary hearing of October 6, 2026, checked November 2026.
Sources
- Voice Over Herald, Unions’ OpenAI brief puts voice actors’ lost work on record, reporting NAVA’s State of Voiceover in 2026 survey and the October 29, 2026 filing: https://www.voiceoverherald.com/unions-openai-fair-use-brief-voice-actors-nava-survey/
- SAG-AFTRA, Statement on Fortnite’s use of AI Darth Vader voice and unfair labor practice filing, May 19, 2025: https://www.sagaftra.org/sag-aftra-statement-fortnites-use-ai-darth-vader-voice-and-ulp-filing
- Slator, SAG-AFTRA voice actors win new protections from AI dubbing, on the ratified 2026 television animation agreements: https://slator.com/sag-aftra-requires-ai-dubbing-consent/
- Reuters via SRN News, Voice actors push back as AI threatens dubbing industry, July 30, 2025: https://srnnews.com/voice-actors-push-back-as-ai-threatens-dubbing-industry/
- Illustrations: Dirty Static
END OF TRANSMISSION
Stay curious. Question everything.
