Gene Wilder AI Voice Clone Made an Oompa-Loompa Cry

The Moment an Oompa-Loompa Broke Down in Tears

Rusty Goffe ends Zoom calls with "Have a scrumdiddlyumptious day!" He says it the way other people say goodbye — naturally, without performance, as if Gene Wilder's linguistic candy-colored world still lives inside him somewhere. That detail matters, because it tells you how deeply the original 1971 production of Willy Wonka & the Chocolate Factory is woven into the people who made it.

Goffe played one of the Oompa-Loompas alongside Wilder in that film. He was there when Wilder, with his particular mix of menace and melancholy, made Willy Wonka into something stranger and more enduring than a children's movie villain. More than five decades later, when Netflix — which acquired the rights to the Wonka universe — used an AI-generated clone of Wilder's voice in a production, Goffe heard it. And he wept.

That image — a man in his seventies, one of the few people alive who actually shared a set with Gene Wilder, undone by the sound of a machine doing an impression of his late colleague — is not just a poignant anecdote. It is evidence. Evidence that what the entertainment industry is doing with AI voice replication carries a human cost that spreadsheets and press releases are unlikely to capture.

How Netflix AI-Cloned Gene Wilder's Voice

Gene Wilder died in August 2016. He left behind decades of film and television work — a rich archive of recordings that, to a modern voice synthesis model, represents training data.

Contemporary AI voice cloning systems, including those built by companies like ElevenLabs and Respeecher, can reconstruct a recognizable vocal signature from a surprisingly small corpus of source material. Research published by voice synthesis labs has demonstrated that high-fidelity clones can be generated from as little as a few minutes of clean audio, with larger archives producing outputs that independent listeners in blind tests struggle to distinguish from the original speaker. The underlying architecture — typically a combination of neural text-to-speech models and voice conversion systems — learns the spectral patterns, prosodic rhythms, and timbral qualities that make a voice identifiable.

Wilder's voice is not obscure. It appears across hours of commercially released film and television, meaning any competent synthesis pipeline would have abundant material to work from. Netflix, which through its acquisition of the Roald Dahl Story Company controls significant Wonka-adjacent intellectual property, used this technical capability to place Wilder's reconstructed voice into new material. The estate's consent, or lack thereof, and the specific terms under which the clone was produced, remain points of legitimate public concern — precisely because the regulatory landscape governing such decisions is still being written.

Why That Reaction Reveals a Real Problem

Grief is not sentimental weakness. When Rusty Goffe heard that voice and cried, he was responding to a violation — not of law, necessarily, but of something harder to codify. The voice he heard sounded like his friend and collaborator. It wasn't.

Media ethicists have a term for this category of harm: epistemic injury. It is the confusion, and sometimes the distress, caused when representation and reality are decoupled without the audience's full understanding. Patrick Lee Plaisance, a media ethics scholar whose work on character and moral agency in journalism and entertainment has been widely cited, has written that the question of AI-generated content is not purely technical but fundamentally relational — it implicates the bonds between creators, audiences, and the cultural artifacts they share.

For someone like Goffe, Wilder was not an abstract cultural property. He was a person. The AI clone performs Wilder's voice without carrying any of the interiority, intention, or consent that made that voice meaningful. The performance is correct in the way a forgery is correct — which is to say, correct in every way that matters least.

This is the problem the tears reveal: a gap between what technology can do and what it should do, inhabited by real people who have no obvious recourse.

The Ethics of Resurrecting Dead Actors With AI

The ethical debate over posthumous digital replication is not new, but it has accelerated dramatically. The 2023 SAG-AFTRA negotiations — the strike that shut down Hollywood for nearly five months — placed AI squarely at the center of labor and rights disputes. The contract that emerged in November 2023 included landmark provisions requiring studios to obtain informed consent from performers before creating digital replicas of their voices or likenesses, and mandating compensation equivalent to what the performer would have received for the original work.

Crucially, those provisions also addressed deceased performers — requiring that estates retain meaningful control over posthumous replication. The language was hard-won, and it reflected a fundamental premise: that a person's voice and likeness are not simply raw material to be industrialized after death.

The Wilder situation tests that premise in a commercially significant context. Wilder's estate controls his rights. But "control" and "consent" are not synonymous. Estates can authorize uses that the deceased performer would have found objectionable, for financial or strategic reasons entirely disconnected from the performer's own values. Laurence Olivier's estate licensed his likeness for a Volkswagen commercial in 2011; many actors and critics found it undignified. Olivier had no say. Neither does Wilder.

Voice actor Tim Friedlander, president of the National Association of Voice Actors, has spoken publicly about the downstream effects of posthumous voice cloning — not just for the deceased, but for living performers whose work is devalued when studios can manufacture a famous voice without paying anyone scale. The economic harm runs parallel to the ethical one.

Hollywood's Ongoing Battle Over Digital Likeness Rights

The SAG-AFTRA 2023 contract was a starting point, not a solution. Its provisions apply to union productions; a significant portion of content development, particularly in streaming and interactive media, operates in jurisdictions or structures where those protections have limited reach. The Writers Guild of America's 2023 agreement similarly addressed AI, prohibiting the use of AI-generated material to replace human writers, but enforcement mechanisms remain nascent.

Intellectual property law has not kept pace. In the United States, rights of publicity — which govern the commercial use of a person's name, voice, and likeness — vary dramatically by state. California's statute extends some protections to deceased performers for up to 70 years post-death; other states offer far less. There is no federal right of publicity statute, meaning that what is prohibited in one jurisdiction may be entirely permissible in another. An AI voice clone produced in a jurisdiction with weak protections and distributed globally via streaming sits in a legal gray zone that courts have barely begun to map.

Several IP attorneys specializing in entertainment law, including those who have written publicly for outlets like The Hollywood Reporter and Variety, have noted that existing copyright frameworks were not designed to address synthetic media. Copyright protects a specific recording, not the voice itself. Training a model on those recordings and generating new outputs occupies contested territory that will likely require legislative intervention to resolve clearly.

What This Moment Should Mean for the Industry

Rusty Goffe's tears should function as an industry-wide prompt — not a veto, but a question that demands a real answer. The question is not whether AI voice cloning is technically impressive. It plainly is. The question is whether the entertainment industry has the institutional will to impose meaningful constraints on its own capabilities.

There are constructive paths. Estates and studios can establish clear public-facing policies about when and how posthumous replication will be used, and what creative standards will apply. Streaming platforms can adopt internal review processes that include human rights assessments before AI-generated content featuring deceased performers is greenlit. Guild agreements can be strengthened to close jurisdictional gaps.

None of this requires abandoning the technology. It requires treating the technology with the seriousness that a crying Oompa-Loompa suggests it warrants.

Gene Wilder spent his career doing something machines cannot — making people feel the specific weight of a human consciousness at work behind a performance. The AI clone can reproduce his vowels. It cannot reproduce what made those vowels matter. That distinction is not sentimental. It is the whole point.

Source: Variety