Chris Williamson's voice-cloning video begins with a deeply unsettling experience: a friend sent him a YouTube narration that sounded so much like Chris that the friend asked whether it was him. Chris traced the sound to an ElevenLabs voice named Archer. The resemblance matters because a recognisable voice can make a fabricated opinion or endorsement sound authentic. But the most important fact in the story is also the least satisfying: the video does not prove that Archer was trained on Chris's recordings.
Chris documents a convincing sound-alike, listener confusion, an ElevenLabs investigation, and Archer's later removal. ElevenLabs told him that a voice actor supplied Archer and reported a low score from its own speaker-similarity test. The video does not provide the training recordings, the actor's identity, an independent test, or a finding that anyone copied Chris. It is a strong account of the identity and trust problem, not proof of a specific act of theft.
Watch Chris Investigate the Voice
The opening fake product pitch is a demonstration, not a real endorsement. Use the chapter links to hear the original audio and Chris's account in context.
The Archer Discovery
At 01:03, Chris plays the MH370 narration that prompted the question. He hears familiar quirks in the North East England accent, not just a generic British tone. By 03:19, the concern has shifted from "who sounds like me?" to "what could a stranger make this voice say?" He worries about ads, opinions and endorsements being wrongly associated with him.
That is why the debate cannot be reduced to a royalty rate. Chris makes his living speaking, but says the deeper cost is the audience's ability to trust that a recording attributed to him really expresses his choices. The video illustrates plausible confusion; it does not establish that every use of Archer was presented as Chris or that the MH370 creator intended to impersonate him.
What Was Tested, and What Was Not
At 07:05, Chris reads ElevenLabs' reply. The company said its voices team had previously liaised with the actor who contributed Archer, was investigating, and had no reason at that point to believe the actor behaved fraudulently. It also reported a speaker-similarity score of 0.28, below a 0.7 threshold it described as a high-confidence match. Chris remains unconvinced by a result that conflicts with what he and others hear.
The transcript appears to refer to an ECAPA-TDNN-style speaker-verification test, but the video does not show the model configuration, audio samples, calibration or full report. The reported number is ElevenLabs' statement in Chris's video, not our independent measurement or a universal probability that the voice is different. A machine's speaker-match score and a listener's sense of resemblance answer related but different questions; neither, on its own, identifies the source recordings.
At 12:49, Chris says Archer was removed from ElevenLabs. He explicitly says he does not know whether that followed his complaint, his public discussion, another decision, or a discovery about the voice. Treat the removal as a reported outcome, not an admission of cloning.
What ElevenLabs Says Its Safeguards Are
It would be misleading to say there are no safeguards. ElevenLabs' current safety page describes voice verification, moderation, a block on high-risk voice cloning and public abuse reporting. Its Voice Library documentation says published user voices must be Professional Voice Clones and are reviewed before sharing. The Voice Library terms require contributors to use their own voice and have the relevant rights. Its use policy links to an unauthorized-use-of-voice report.
Those are the company's published rules and controls, not an independent audit of this particular voice or proof that every sound-alike will be detected. Chris's case exposes a harder question: what clear, appealable evidence can a person obtain when a platform's provenance account and the person's own experience disagree?
The Legal Examples Have Narrow Facts
Chris speaks with a lawyer at 09:24 and points to two US advertising cases. In Midler v. Ford, the court considered evidence that advertisers deliberately hired a singer to imitate Bette Midler after she declined to appear in a commercial. Waits v. Frito-Lay likewise concerned a deliberately imitative commercial performance. Those decisions show that, under the California law applied there, deliberate commercial imitation of a distinctive, widely known voice can matter even without using the person's original recording.
They do not decide Chris's situation. The video's unresolved questions include whether Archer was deliberately made to evoke him, how it was trained, who used it and for what, and which jurisdiction's law would apply. Similarity alone is not the same as appropriation. The lawyer in the video recommends preserving evidence and consulting a litigator about a specific claim; this article is not legal advice.
A Practical Response to a Suspected Sound-Alike
If you encounter a voice that seems to imitate you, preserve the original URL, audio, date, voice listing, context, and any messages from people who believed it was you. Record exactly what you know and separate it from what you infer. Use the platform's unauthorized-voice reporting route if applicable, and ask for a written explanation of the review outcome. For commercially significant or harmful uses, take that evidence to a lawyer in the relevant jurisdiction before making a public accusation.
For listeners, the same caution runs in reverse: a familiar voice is no longer sufficient proof of authorship. If a recording asks for money, credentials or an unusual action, verify it through a channel you already trust. The US FTC advises calling a person back on a known number when a voice-cloning scam appears to involve someone you know. Chris's case is not presented as that kind of scam, but it makes the verification habit easier to understand.
Video Chapters
| Time | Topic |
|---|---|
| 00:00 | Opening: an AI voice that sounds like Chris |
| 01:03 | A friend hears a familiar voice in an MH370 video |
| 03:19 | Archer and the fear of false endorsements |
| 05:15 | Why voice identity is more than a licensing fee |
| 07:05 | ElevenLabs responds and describes its similarity test |
| 09:24 | The US voice-imitation cases and a lawyer's view |
| 12:49 | Archer is removed; the reason remains unknown |
Sources and Limits
Primary account: Chris Williamson's video and the supplied transcript. The email and test result are shown through Chris's account; we have not seen the underlying files or conducted a forensic comparison. Platform rules: ElevenLabs' safety overview, Voice Library guide, terms and use policy. Technical and legal context: the ECAPA-TDNN paper and the Midler and Waits opinions. Public sources checked 23 October 2026. None independently establishes who provided Archer's training audio or why it was removed.