At the age of 34, while pregnant with her second child, Sarah Ezekiel from north London was diagnosed with Motor Neurone Disease (MND). The condition progressively damages parts of the nervous system and often weakens the tongue, mouth, and throat muscles, leaving many sufferers unable to speak.
Following her diagnosis, Ezekiel lost the ability to use her natural voice. She turned to voice-generating technology and computers to communicate, but the artificial voice sounded nothing like her own. Despite these challenges, she continued her career as an artist, creating images using a computer cursor. However, her children, Aviva and Eric, grew up never hearing how their mother had once spoken.
Challenges of Voice Preservation
In recent years, experts have explored ways to recreate computerised versions of a person’s original voice. Traditionally, this process required long, high-quality recordings, which often produced voices that were flat and monotone. “We originally asked Sarah for an hour’s worth of audio,” said Simon Poole of the UK medical communication company Smartbox.
As reported by france24, doctors today encourage patients with conditions like MND to record their voices early, preserving a key part of their identity. Yet, in the pre-smartphone era, suitable recordings were rare. Ezekiel could find only one very short, poor-quality clip—a muffled eight-second home video from the 1990s with background television noise. “My heart sank when I heard it,” Poole admitted.
Turning to AI for a Solution
Poole then turned to AI technology developed by New York-based company ElevenLabs. One tool isolated Sarah’s voice sample from the noisy clip, while another—trained on real voices—filled in the missing elements. The final product was remarkably close to her original voice, complete with her London accent and even the slight lisp she once disliked.
“I sent her samples, and she emailed back saying she nearly cried when she heard it,” Poole recalled. “She played it to a friend who knew her before she lost her voice, and it was like having her voice back.”
Restoring Identity Through Technology
According to the UK’s Motor Neurone Disease Association, eight in ten people with MND face voice difficulties. Current computer-generated voices often sound robotic, lacking pitch, tone, and emotion. Poole highlighted the breakthrough: “The real advance with this new AI technology is that the voices are human, expressive, and restore the humanity that was missing in older computerised voices.”
He added that personalising a voice helps preserve identity. “If you lose your voice later in life, being able to speak in your own voice is incredibly important. It feels far more authentic than using an off-the-shelf computer voice.”
A Voice Returned
For Sarah Ezekiel, this innovation means more than technology—it means regaining a vital part of herself. After 25 years, she can once again be heard in her own voice, bridging the silence MND once imposed.