Margot Callahan, of Highland Park, is providing her voice for a stranger – literally. She’s one of thousands of people who have donated their voices to people with vocal disabilities, caused by a range of factors such as a stroke, cerebral palsy or Lou Gehrig’s Disease.
VocaliD is collecting those voices and using them for voice software devices. The goal is to provide a more accurate representation of the actual person’s voice, rather than being stuck with a robotic generic one.
Callahan said she was inspired to donate her voice after hearing about VocaliD on NPR.
“My hope is that someone who has not been able to express themselves, maybe in years, will finally be able to say something in a voice that is like their own and in a way that is how they feel,” she said. “I think that would be a most wonderful thing.”
Rupal Patel, founder of VocaliD, said the idea of the company came to her when she observed a young girl and a grown man having a conversation with different kinds of speech assistance technology. Despite their age and gender differences, they spoke through the same type of generic voice; think Stephen Hawking.
“Our voices are so much tied to who we are, what we like to do and our personalities,” Patel said. “And people who can't speak have been denied that.”
Patel said the incident made her wonder if there was a way to more accurately capture the character of those who may be vocally-impaired.
“Even though they can’t speak clearly, they still have a natural voice like you and I do,” Patel said. “And maybe there's a way to build a machine that if it’s prosthesis for them, it could actually somehow represent them and not just be this box that they talk through.”
Patel said VocaliD began as a research project in Patel’s lab at Northeastern University. She and a team of people successfully developed technology that fuses that remaining aspect of someone’s voice with the recordings from a speech donor. The end result is a unique voice which is compatible with several of the devices and apps the recipients already speak through.
Callahan just finished her recordings for VocaliD’s voice bank after five months of recording roughly 100 to 200 sentences each day. She recorded a total of 3,487 phrases, which is what VocaliD requires to create a high-quality voice. She said she recognized some of the sentences from classic books, such as The Wizard of Oz and Little Women.
A new voice through VocaliD has an initial fee of $1,249 and then a $240 annual subscription to continue to tune the voice, according to Patel. She said this is used to update the voice as the user gets older or to correct mispronunciations.
Patel said there are about 20,000 people in the voice donor program, with about 12,000 currently recording their voices. Back in December, the first synthetic voices were created for seven recipients. The company is now preparing 80 voices for a second launch later this year.
