Voice
Actors Push Back as AI-Generated Voices Threaten Dubbing
Industry
The dubbing booth has always been where invisible labour makes
global storytelling possible. The voice actors who re-perform dialogue in languages
other than the original bring characters to life for audiences who would
otherwise need subtitles to follow films, series, games, and animations.
Their work is skilled, economically significant, and almost entirely
uncredited in the cultural conversation about how entertainment crosses
linguistic borders. Now that work is under threat from AI voice synthesis,
and the performers doing it are fighting back.
The pushback from voice actors in Europe and beyond is not simply
a labour dispute about wages or contracts, though it is that too. It is a
dispute about identity, consent, and the question of whether a performer’s
voice, the most intimate of their professional tools, can be replicated,
sold, and deployed without their knowledge or permission. The answer that the
AI industry has implicitly provided so far is yes, and many performers and
their unions have decided that answer is not acceptable.
What AI Voice Synthesis Actually Does
Modern AI voice synthesis systems can clone a voice from a
relatively small sample of recorded audio. Given training data, which may be
drawn from commercially available recordings made for entirely different purposes,
a synthesis model can generate new speech in that voice delivering any text,
in any language, with controllable prosody and emotional colouring. The
output quality of current systems is sufficient to pass casual listening
scrutiny, and it is improving rapidly.
For dubbing studios, the commercial proposition is
straightforward: a synthesised voice that sounds sufficiently like a
recognised actor costs a fraction of what that actor charges per session. For
global streaming platforms distributing content across dozens of language
markets simultaneously, the cost differential across a full catalogue is
enormous. The economic incentive to use AI voice synthesis in dubbing is
powerful, and it is already being acted on.
Wired
reported on the resolution of the SAG-AFTRA video game performers’
strike, which produced new contract language requiring explicit consent
before AI-generated replications of actors’ voices can be used. That
settlement was hard-won and specific to the games sector. It does not cover
film and television dubbing, where the regulatory landscape remains largely
open.
The European Response
European voice actors have been particularly vocal in demanding
legislative protection. In France, Germany, Italy, and beyond, performers’
unions are pushing for EU-wide rules that would require explicit consent
before any AI system can be trained on a performer’s voice, mandate
compensation when synthesised voice is used commercially, and require
disclosure to audiences when dubbing has been produced by AI rather than
human performers.
The calls from dubbing professionals reported by Reuters
and others centre on three principles: consent, compensation, and
traceability. Consent means performers must actively agree before their vocal
characteristics are captured or replicated. Compensation means the economic
value generated by synthesised voice must flow back to the performer whose
voice provided the training data. Traceability means audiences and
commissioning bodies must be able to determine whether a dubbed performance
was created by a human or generated by AI.
The EU AI Act contains provisions relevant to biometric data and
manipulative AI, but its application to voice synthesis in entertainment is
not yet settled. The performers’ unions are lobbying for sector-specific
regulation that addresses the particular character of voice acting as an art
form and a livelihood.
Identity Theft and the Personal Dimension
For many voice actors, the concern extends beyond economics into
something more personal. A voice is not merely a professional instrument. It
is an expression of identity, shaped by a lifetime of experience and
training, and its unauthorised use in contexts the performer has not approved
carries a violation that financial compensation alone does not fully
address.
A UNLV
essay examining voice actors and AI replication documents
performers describing the experience of hearing their vocal likeness used in
content they did not perform, sometimes in contexts that conflict with their
values, public identity, or professional standards. The emotional toll is
significant, and the legal remedies are currently limited.
An arXiv paper on biometric
replication and performer rights identifies voice synthesis as
raising systemic risks to performer privacy, agency, and dignity that
existing intellectual property frameworks were not designed to address. The
paper calls for new legal categories that treat a performer’s voice as a
protectable biometric asset rather than as content that enters the public
domain through commercial performance.
The Broader Creative Industry Context
Voice acting is one front in a wider conflict between AI
capabilities and creative labour. As explored in AI
in Theatre, AI tools are being applied across the performing arts
in ways that raise questions about authorship, credit, and the economic
relationship between human creativity and machine assistance. Voice acting is
a particularly acute case because the technology is mature enough to produce
commercially deployable output right now, and the economic pressure to use it
is immediate rather than theoretical.
The alignment between what AI can do technically and what
production economics incentivise is a recurring pattern. As discussed in
The
Forgotten Accent, the communities most affected by AI’s linguistic
capabilities and limitations are often those with the least power to shape
how those capabilities are deployed. Voice actors in minority language
dubbing markets are particularly vulnerable: their work serves audiences that
are commercially less valuable to global streaming platforms, making the
economics of AI substitution even more favourable from a corporate
perspective.
What Protection Would Look Like
Effective protection for voice actors from unauthorised AI
replication requires action at multiple levels. At the contractual level,
industry standard agreements need to include explicit prohibitions on the use
of recordings for voice synthesis training without separate consent and
compensation. At the legislative level, biometric data protection frameworks
need to be extended and clarified to cover vocal characteristics used for
synthesis. At the technical level, watermarking and provenance systems for
AI-generated audio need to become standard so that synthesised voice can be
identified and attributed.
None of these protections yet exists at scale. The industry is
moving faster than the regulatory frameworks designed to govern it, and the
window for establishing meaningful protections before AI voice synthesis
becomes fully embedded in production pipelines is closing. The voice actors
demanding action now are not asking to stop the technology. They are asking
for the same basic conditions that have governed creative labour in every
other technological transition: consent, compensation, and
credit.
Those are not unreasonable demands. Whether the industry and its
regulators will meet them before the transition is complete is the question
the next two years will answer.
The stakes for the creative economy are significant. Dubbing is
not only a livelihood for tens of thousands of performers across Europe and
beyond; it is one of the primary mechanisms through which audio-visual culture
crosses language barriers and reaches global audiences in forms that feel
genuinely local. A dubbing industry hollowed out by AI voice synthesis would
produce cheaper localised content, but not better or more culturally resonant
content. The emotional craft that makes great dubbing work cannot be
generated from training data. As the digital literacy concerns raised in
Google’s
AI Mode Transforms Search illustrate, audiences are not always well
placed to detect the difference between genuine human performance and a
convincing simulation. That is precisely why the performers themselves, and
the unions representing them, cannot afford to wait for public awareness to
catch up.
About the Author
Stuart Kerr is the Technology Correspondent for LiveAIWire,
covering artificial intelligence, ethics, and the ways technology is
reshaping everyday life.