LONDON: A British woman has regained her original speaking voice after twenty five years of silence caused by motor neurone disease, thanks to groundbreaking artificial intelligence technology.
Sarah Ezekiel lost her ability to speak when diagnosed with MND at age thirty four while pregnant with her second child a quarter century ago.
The progressive neurological condition damages the nervous system and can weaken tongue, mouth and throat muscles, often resulting in complete speech loss.
Ezekiel had previously used computer generated voice technology that sounded completely different from her natural voice.
Her children grew up never hearing their mother’s original speaking voice until this technological breakthrough.
Modern voice recreation techniques typically require high quality recordings of at least one hour duration according to experts.
Simon Poole of medical communication company Smartbox explained that traditional methods often produced “very flat and monotone” results despite longer recordings.
The company initially requested an hour of audio from Ezekiel but faced disappointment when only one poor quality clip could be found.
An eight second home video clip from the nineteen nineties provided the only available sample of Ezekiel’s original voice.
The muffled recording contained significant background noise from a television, making the task particularly challenging.
Poole utilized AI technology from New York based company ElevenLabs that specializes in voice recreation from minimal samples.
One AI tool isolated the voice sample while another, trained on real human voices, reconstructed the complete vocal pattern.
The resulting voice perfectly captured Ezekiel’s original London accent and even reproduced her slight lisp.
Ezekiel reported nearly crying when she first heard the AI generated version of her own voice after twenty five years.
A friend who knew Ezekiel before her voice loss confirmed the remarkable accuracy of the recreated voice.
According to the UK Motor Neurone Disease Association, eighty percent of sufferers experience voice difficulties after diagnosis.
Current computer generated voices often sound “quite robotic” in their timing, pitch and tone according to the association.
Poole emphasized that this new AI technology brings “humanity back into the voice that previously sounded a bit computerised”.
He stressed the importance of voice personalization for preserving individual identity after acquired conditions.
“Being able to speak using your original voice is really quite important, rather than using some off the shelf voice,“ Poole concluded. – AFP