Woman with neurodegenerative disease recovers voice with AI

Eric and Aviva grew up without hearing their mother's voice, as she was affected by a neurodegenerative disease, but thanks to an old audio recording of hers, processed using artificial intelligence (AI), they can now hear her.

Twenty-five years ago, while expecting her second child, Sarah Ezekiel discovered she had a degenerative disease that can cause some patients to completely lose their ability to speak. This was her case.

+ Japanese city wants to limit smartphone use to two hours a day

After the impact of the diagnosis, she was able to use a computer and speech synthesis technology to communicate, but with a robotic, completely dehumanized voice.

Her children, Aviva and Eric, therefore grew up without knowing the sound of their mother's voice.

Until the day the family contacted the British company Smartbox Assistive Technology, which develops communication tools, especially with artificial intelligence, for people with disabilities.

The goal was to recover Sarah's true voice.

To this end, the company asked the family to provide old recordings of the woman, warning that they should be sufficiently long and of good quality.

People suffering from degenerative diseases are currently being encouraged to record their voices as soon as possible so they can use AI.

But before the digital age and smartphones, it was much rarer to have adequate recordings.

Sarah Ezekiel was only able to recover a fragment of her voice. Taken from a family video from the 1990s, it was only eight seconds long and of poor quality, as the sound of her voice was interfered with by the audio from a television program.

But the British company didn't give up. It began by isolating Sarah's voice before using generative AI.

The results exceeded expectations. The technology even managed to reproduce Sarah's slight lisp. "I sent her samples, and she emailed me back saying she almost cried when she heard them," Simon Poole of Smartbox told AFP.

"She had a friend listen to it who knew her before she lost her voice, and that friend said it sounded like Sarah had gotten her voice back," a delighted Poole told AFP.

For him, the real breakthrough lies in the fact that, with AI, “the voices are truly human and expressive, and they restore that humanity that until now was missing in voices that were a little too synthetic.”

Ultimately, this allows “preserving the person’s identity,” he summarized.