There are many forms of adaptive technology that give voices to non-verbal people. Famously, theoretical physicist Stephen Hawking is able to talk by moving only his cheek. He has an infrared sensor on his glasses that can pick up on muscle movement; this is connected to a computer with advanced word prediction software. Once he has written something, he can send it to a speech synthesizer, which reads the words aloud (1).
His computerized voice has a distinctly mechanical quality, which his fans greatly enjoy. Reportedly, so does Hawking. Although he embraces technological updates to his communication system, he has been adamant that the sound of his voice – which was developed in the early 80s – remain the same (2).
However, other nonverbal people feel differently. “My voice isn’t me. It sounds like a robot. I want it to be me,” explains Maeve, a young girl with cerebral palsy (3).
Fortunately for Maeve, she was able to work with a cutting-edge technology company called VocaliD to make this dream a reality.
CEO and co-founder Rupal Patel saw a need for individualized voices in text-to-speech technology when she observed a young girl, not unlike Maeve, speaking with an assistive device that made her sound like an adult male (4).
“We wouldn’t dream of fitting a little girl with the prosthetic limb of a grown man,” her website notes (3). “So, why then the same prosthetic voice?”
VocaliD actually creates a unique voice for each client. They have a global repository called The Human Voicebank, with vocal samples from more than 20,000 speakers from all around the world. These people can record themselves speaking from the comfort of their own homes. VocaliD stresses that one does not need to have a flawless voice to become a vocal donor. Some clients even request voices with a lisp or a stutter (5).
The voice donors are not paid, but many find it to be a highly rewarding experience. “I’m recording my voice. I got to ‘I love you’ and burst into tears,” says a voicebank contributor identified as Chris K. “Realizing that I could be helping someone to say that phrase really drove home the importance of the work you’re doing. What an honor to be a part of it!”
VocaliD’s clients contribute three seconds of their own sound (this does not need to be coherent speech – any vocalizations will do), and VocaliD blends aspects of their “vocal DNA” with that of a donor from the voicebank. It isn’t cheap ($1,499), but the voice created is compatible with a variety of assistive devices, and can be used for a lifetime. They can even adapt as the client ages.
In addition to creating blended, unique voices for nonverbal clients, VocaliD also works with people who have degenerative diseases like ALS to preserve their voices before they are lost. The process is similar, except that they are essentially their own vocal donor. In order to do this, they will need to submit 5-7 hours of recorded speech.
The voices created by VocaliD still sound somewhat mechanized, but certainly seem more natural than the generic text-to-speech options. Their clients agree. Of her new voice, Maeve says, “It sounds more like me” (4).
To watch Maeve and her best friend, who is also nonverbal, communicate using their new VocaliD voices, click here.