Your Voice Stays Yours After Death, Japan's Actors Union Insists

Japan's actors union has told the AI industry that no one may clone a performer's voice without consent, and that, with narrow exceptions, death does not change the answer.

The Japan Actors Union, which says about 2,600 actors and voice actors belong to it, announced its Guidelines for the Use of Generative AI Related to Voice on October 15 ^[1][5]. They are an industry proposal, not a law. No company is bound by them.

But they set out, in unusual detail, what a performer's union wants before AI voice replication becomes routine in animation, games, dubbing and other audio work. They also arrive from a country whose voice-acting industry, the world of the seiyuu, is among the most visible on earth.

Consent at every step

The guidelines go well beyond a ban on stolen voices. They call for performers to be told what an AI voice will be used for, and to give prior consent at three separate points: when a voice is used to train a model, when speech is generated, and when the result is published or used ^[1][5].

That matters because blanket permission is the easy route for a studio. A single signature covering everything a company might later want would not satisfy a framework that asks for a fresh yes at each stage.

The union also treats AI voice use as a new kind of performance, one that deserves pay of its own, separate from ordinary performance fees. The guidelines name payment for training or development, for use of the generated voice, and for multilingual use ^[1][3].

The multilingual provision reaches into dubbing. If a performance is translated into other languages by AI, the guidelines call for new permission and additional compensation, and say the artistic integrity of the original performance should be preserved ^[1][3].

In its public statement the union said technological progress will continue, but that "human expression" and respect for rights must remain central ^[2].

The dead

The deceased-performer rule is the most striking part. The guidelines set a general prohibition on recreating a dead performer's voice with AI ^[3].

It is a strong rule but not an absolute one. The exceptions include projects the performer approved while alive, and sequels to series already in production when the performer died ^[3].

Even then, use is meant to be limited. It requires written consent from the performer's heirs and from the former agency or organization, and payment goes to the heirs ^[3].

The design is worth noticing. A voice is the one part of an actor that outlives them in recordings, and a model trained on those recordings could keep a character speaking indefinitely. The union has drawn the line at the performer's own wishes in life, then added the family and the agency as gatekeepers.

A gap in the law

The union says Japan has no comprehensive law that explicitly protects an individual's voice ^[4][5]. It describes the guidelines as a response to unauthorized collection of performers' recordings and to that legal gap ^[1].

Some protection exists. According to ITmedia, government guidance has said that voice use may fall under publicity rights, the legal claim that a person controls the commercial use of their identity.

A Tokyo court has also recognized that a voice may qualify for that protection in a case involving the actor Kenjiro Tsuda. The court nonetheless rejected his request to have the videos removed ^[4].

That result shows why the union is not content to wait for the courts. Recognition of a right in principle, without a remedy in the case at hand, leaves a performer little better off. The union says it intends to encourage industry adoption of the guidelines and to press the government to legislate protections for voice rights ^[4][5].

Beyond Japan

The union says it hopes the guidelines can help shape a global standard for the film and audio-content industries ^[1]. That is an ambition, not a development. No overseas union or company has been reported to have adopted the guidelines, and no international negotiations have been reported.

Still, the document gives other unions something concrete to point at. Most debates over AI and performers' rights are fought over a single point, such as whether a studio may scan a face or clone a voice. This framework covers training, generation, publication, pay, translation and death in one place.

It is also candid about what it cannot do. A voluntary code works only if the people it addresses agree to sign contracts that follow it. Whether Japanese production companies, AI developers and platforms do so is the open question.

So is how the union would monitor compliance, particularly for recordings and voice models that already exist.

The next tests are practical. Watch for the first production to cite the guidelines in a contract, for any move by the government on voice-rights legislation, and for a reply from outside Japan.