Voice Cloning Preserves Identity for People Losing Speech to MND

Gabi Leibowitz described ElevenLabs’ work with people with motor neuron disease as a use case the company “never could have imagined” when it first appeared in 2024. The need, as she framed it, is not simply that people with MND may lose speech. It is that many lose the physical ability to express themselves while their minds “often remain fully intact”: they still think, feel, and create, but progressively lose the muscular control needed to walk, use their hands, eat, and speak.

For people who cannot speak, assistive technology has long provided a functional bridge. A person types a message; a device speaks it aloud. But Leibowitz drew a distinction between being able to get “basic needs met” and being “truly heard.” The historical output, in her description, has often been a generic robotic text-to-speech voice. It communicates content while stripping out much of the person.

That is the gap ElevenLabs says voice cloning can address. Leibowitz emphasized that the company’s process does not treat recognizable vocal traits as defects. A stutter, a lisp, a laugh, or a regional dialect are not “imperfections to be removed,” she said, but part of the person. The aim is not a polished neutral voice. It is continuity: preserving what makes a speaker sound like themselves.

The same claim extends to language. Leibowitz said vocal identity is not limited to one language, because many people move between languages at home, at work, and with friends. ElevenLabs’ cloned voices can be used in more than 70 languages, allowing people to switch between them “in a voice that feels authentically theirs.”

The uses Leibowitz listed were deliberately broader than medical communication. She said people with voice loss using ElevenLabs were staying in jobs longer, performing standup comedy, writing and producing audiobooks, renewing vows, and reading stories to grandchildren. The company’s Impact program was presented as the vehicle for that work. A stage slide displayed the program name alongside partner names including Witness Films London, the Motor Neurone Disease Association, St George’s Chapel Windsor Castle, and speech and language therapist Richard Cave.

Leibowitz said ElevenLabs has committed to “giving back one million voices” through the program: free lifetime access for anyone with permanent voice loss, including from MND, head and neck cancer, stroke, and cerebral palsy, as well as for the speech therapists supporting them.

She also said the program had already enabled more than 10,000 free voices. The scale mattered in her framing, but the center of the presentation was not the number. It was the story of Irene Perrin, a history teacher from 1973 and a head teacher for 31 years, living with MND, whose voice had been cloned and who was still using it in public life.

Irene Perrin located her sense of self in history, teaching, and public explanation. At St George’s Chapel Windsor Castle, she described feeling “calm and serenity” as well as “a love of history.” Everybody, she said, needs to know where they come from; continuity matters in life; helping people understand their place in history gives her joy.

That continuity had been bound up with her voice for decades. Perrin said she started as a history teacher in 1973 and “absolutely loved it.” Her central aim as a teacher, she said, was to believe in the pupils. She also described her voice as a professional instrument: “I have always had a voice that had command. I learned to use it in different ways. So my voice has been my life.”

MND changed that. Perrin said her current voice felt alien to her: “This is not me.” She was diagnosed in May 2024, after a life of fitness, and the first symptom she noticed was deterioration in her speech. She began slurring her words and struggling to enunciate certain words and phrases.

Her speech therapist, Richard Cave, proposed that she bank her voice. Perrin called that “the best advice possible.” Cave then told her about ElevenLabs and the possibility of reproducing her voice using AI. Perrin said she “jumped at the opportunity” and was grateful to hear what she used to sound like.

Martin Perrin called MND “a very cruel thing.” After 50 years of marriage, he said, hearing what Irene had sounded like again was “very emotional.” Describing their relationship, he called them soulmates and said the most difficult part was knowing there was nothing he could do about the disease. “I don’t know how I’m going to cope with her,” he said, “and I don’t know how I’m going to cope without her.”

Perrin’s own phrasing held the contradiction at the center of the technology’s role. She could not undo the disease. “I can’t rewrite it,” she said. But she called it ironic that her voice had been “taken away” and, in a way, “given back, because of the app.”

Perrin’s most concrete example was her continued volunteering as a steward at St George’s Chapel. She began stewarding there in 2007 and said she loves interacting with visitors. Without the app, she said, there was “no way on this planet” she could still do it.

The workflow she described was practical rather than symbolic. She wears the device around her neck when she is on duty. When a visitor asks a question, she types into the app or selects an answer she has prepared, presses a button, and the app speaks in her voice.

She has prepared extensively. Perrin said she had typed “every possible answer” to “every possible question” she could be asked about the chapel’s history into the app. If a visitor asks why Henry VIII is buried in the middle of the floor in the quire, she can press a button and her “old head teacher voice” comes through with the answer.

One guided-history answer captured the kind of speech the app was enabling. Perrin’s voice explained that if visitors stand in the nave and look up, they can see the stone-vaulted ceiling and the decorated bosses that conceal the joints where the ribs meet. There are 463 such bosses, carved in stone. The example is not an emergency message, a yes-or-no exchange, or a request for care. It is the sort of fluent, domain-specific explanation that had defined Perrin’s working life.

Visitors, she said, respond strongly. She has had tears, hugs, and praise for continuing despite her difficulties. A colleague at the chapel cried when Perrin played the cloned voice. “The real Irene was back in the room,” Perrin said.

The technology did not appear as frictionless. When Leibowitz asked Perrin what people should know about her story, the voice clone did not immediately respond. Leibowitz acknowledged the failure plainly — “Not working. All right. We’re in for it” — and Perrin continued. Perrin still described the app as enabling work she otherwise could not do.

To continue at the chapel, she said, “means the world” to her. As a teacher and head teacher for 31 years, her voice had been her lifeline. To have it stolen was “so very cruel.” But she rejected the conclusion that losing her natural speech meant she should withdraw. “I could simply give up,” she said, “but to me, what little life I have left is worth living.”

Irene Perrin argued for voice banking as more than convenience. She described speech loss as a threat to self-esteem, social standing, and participation. It is easy, she said, to give up on life when identity and vocal communication feel stolen. But “it doesn’t have to be like that.” With a voice app, she said, people can still contribute in employment, volunteering, leisure pursuits, and personal life.

Her advice to employers, friends, and family was direct: see beyond the disability and value the person as before. That was not only a plea for kindness. Perrin described a specific form of social misreading that can come with impaired speech. People may assume someone with speech difficulty has dementia or “had a few beers at the pub,” and then treat them “appallingly.”

That is where she placed “AI for good.” True inclusion, for Perrin, would mean people with speech difficulties or no voice are no longer marginalized, and are represented in every sphere of life on equal footing with able-bodied people. No doors would be closed to them. No one would judge them because of disability. In an ideal world, she said, AI used for good would help dispel the myths that attach to impaired speech.

The app, as Perrin described it, does not cure MND and does not restore spontaneous natural speech. It gives her a way to answer, teach, and be recognized in situations where the content and recognizability of a voice matter. When she says the app lets her continue to feel valued and contribute meaningfully, she is describing a social function as much as a technical one.

Gabi Leibowitz echoed that framing in her closing response, saying ElevenLabs would continue striving for “true AI for good” that takes into account the whole person and empowers people to keep thriving and sharing their light. Perrin’s description of inclusion was more specific: not being misread, excluded, patronized, or assumed cognitively impaired because the mechanics of speech have changed.

Irene Perrin closed with gratitude for ElevenLabs and for the professional team behind the work, but she also named a limitation. Voice cloning, she said, is “a really wonderful start,” not the endpoint. The next challenge she wants addressed is spontaneous conversation when speech is severely impaired.

Her point was that there appears to be “no technology support for understanding spontaneous conversation with severely impaired speech.” A solution to that, she said, would be “a dream come true.” It would also need to work in any environment.

The gap follows from the way Perrin described her current use of the app. Her chapel work is possible partly because she can prepare answers to recurring historical questions and trigger them when needed. That is powerful for formal explanation, guided volunteering, and anticipated interactions. Spontaneous conversation would ask something different: support in live, varied settings where the exchange is not limited to prewritten text or predictable questions.

Her final line returned to the language of the head teacher whose voice the app had helped preserve. AI voice cloning, she said, “shows great promise, and capable of even greater achievements.” That was both praise and an assignment.