Ron Brady was 52 years old when ALS, short for amyotrophic lateral sclerosis, commonly known as Lou Gehrig’s disease, is a neurodegenerative disorder that eventually leaves most people unable to speak, walk, or It can cause you to lose your ability to breathe.
Now, at 55, he can’t swallow food, and it’s getting harder and harder to brush his teeth and get dressed. It’s unclear to this point.
But he hasn’t lost his voice.
The reason is that he preserved his voice with a company called Voice Keeper. Voice Keeper is one of several companies that use artificial intelligence to “bank” people’s voices while they are still able to speak and recreate those voices for text-to-speech software. is.
Voice banking used to be expensive and time-consuming, but AI has made it more accessible to people with conditions that can affect their ability to speak, such as ALS, laryngeal cancer, cerebral palsy, and Parkinson’s disease. rice field.
Patients say that having a computer-generated voice that sounds like their true voice has increased their confidence and made them more connected to the world around them.
Brady’s synthesized voice is not a perfect match. His speech was already impaired when he recorded himself. But it has the same relaxed, deep tone that he jokingly calls “gentle”.
“My favorite words are the corny dad comments that make my wife and adult kids laugh,” he said.
For Brady, regaining his voice was like regaining a part of himself. A school administrator who commanded the room with confidence, a gregarious and talkative father, and the first college graduate in his family, with a neutral American accent that was very different from the Caribbean. . The accent of his immigrant parents.
Voice banking is booming, especially among ALS patients, due to the use of artificial intelligence. In 2017, Team Gleason Foundation, a nonprofit that funds voice banking for ALS patients, received 172 requests for the service. In 2022, we received over 1,200 requests. An average of 5,000 people in the United States are diagnosed with her ALS each year.
Capturing human speech is incredibly complex. It used to be necessary to record 1,000 to 6,000 sentences to capture all possible sounds in a language. This process typically takes 8-30 hours. These recorded sounds are put into a database and software rearranges the sounds to form words and phrases.
This method is known as unit selection, and the results were “choppy,” said Tim Bunnell, director of the Nemours Center for Pediatric Auditory and Speech Sciences.
“It’s understandable, but very jarring,” Bunnell said. “Our troop selection voice doesn’t sound as good as a human voice.”
His lab has moved from old methods of speech synthesis to new methods such as using artificial intelligence.
To create digital voices, AI software analyzes human voice samples and quickly scans large databases to find people with similar speaking styles. Find patterns in how speech sounds and create digital voices that match individual speakers.
Most companies today only need a few hundred sentences to get enough data. But some companies, like the Acapela Group in partnership with the Team Gleason Foundation, have algorithms that can create speech from as little as 50 sentences.
With the use of AI, voice banking has also become more affordable. Acapela Group charged him $3,000 when the company relied on unit selection, but with AI, the cost is now $999. Other companies offer this service for as low as $300. Voice banking isn’t covered by insurance, but most companies won’t charge unless they start using synthetic voice.
John M. Costello, who has worked with thousands of patients as director of the Augmented Communications Program at Boston Children’s Hospital, encourages patients to work with speech-language pathologists to find the product that best fits their abilities and needs. Recommended. He found that patients with realistic voices had more meaningful connections with their loved ones.
“Individual voice is very important to our relationship,” he said. “There is a psychological response.”
Studies show that hearing your mother’s voice releases the same level of oxytocin as a mother’s hug. Oxytocin is a social bonding hormone associated with lowering levels of stress and anxiety. Another study found that listening to one’s own voice enhanced self-agency.
other language
Anna Paula Pereira Hülle Mateus, 51, of Lafayette, California, was convinced by the ease of use of the new technology to try it out. When she was diagnosed with ALS in July 2022, she hesitated to spend her energy focusing on what she could lose. She changed her mind when she was told by her doctor that her banking would take about an hour and that she would be able to have her voice anytime.
“I feel like my speech is getting worse and worse, so I’m very happy to do it now,” she said.
However, the a cappella group she was with does not offer services in her native language, Portuguese. To offer Voice Her banking in different languages, companies would have to develop separate algorithms for each language.
The fact that Pereira Hulle Mateus was unable to keep her voice in Portuguese makes her sad because her mother and many of her close family and friends only speak Portuguese. is.
In English, I sometimes have a hard time finding words to express myself. But when she speaks Portuguese, her voice rises and falls like musical notes.
Pereira Hulle Mateus’ synthetic English voice is a little flat, but still captures her distinctive Brazilian accent.
But she’s not going to listen until the moment she needs to use her synthesized voice.
Standardization
Each company has its own method of capturing audio, using different sentences and algorithms. Blair Casey, executive of the Team Gleason Foundation, a non-profit that helps ALS patients, says if someone loses the ability to speak after entrusting his voice to one of his companies, he’ll be stuck with that company. said it is possible.
Casey has asked companies to create a standardized set of phrases that any algorithm can use to help customers compare their purchases. He also asks companies to provide customers with original recordings for future use by other companies.
“If something better comes out, wouldn’t you like to try it?” he asked. “And if you can’t access these phrase sets, you can’t.”
Brian Wallach, 42, a prominent ALS activist and former federal prosecutor in Illinois, was diagnosed with ALS when he was 37. Over the years, his voice has changed from powerful and clear to a mumbling murmur.
He said the first time he played the synthetic voice to his family, it was so accurate that his wife was moved to tears. Meanwhile, his youngest daughter, who had never heard his voice before she had ALS, asked, “Is that you, Daddy?”
“I said back to her, ‘Yes. My voice has changed a lot, but this is what I used to sound like,'” he said.
He likes synthetic voices, but he doesn’t pronounce his wife’s name, Sandra, correctly. Even the synthesized voice fails to express the emotions he wants to convey when speaking to his two young daughters.
Typing out what he wants to say in synthetic voice is a slow process because his hand muscles are weakened.
Due to the limitations of technology, Wallach tends to use synthetic voices only when in public, at larger gatherings with friends, or when he is too tired to speak. can understand most of the
“Have Your Own Voice”
When ALS sufferers lose the ability to use their hands, they have to use their eyes to type, further slowing speech.
This was the case with Ruth Blanton of Rogers, Arkansas. She was diagnosed with her ALS in March of 2021 and she lost her ability to speak by Christmas of that year.
She spoke up shortly after her diagnosis, but the company she worked for used unit selection technology. She spent about a month recording her 3,000 sentences, but was unsatisfied with the final result.
So she was stuck with using a generic voice with an American accent called Microsoft’s “Heather.” Her voice, however, failed to capture her mild British accent from Ormskirk, England. They called it Posh.
In the voice of “Heather,” Ruth, a pragmatic and strong-willed person who was the CEO of a nonprofit that helped struggling families, began to withdraw into her shell, said her husband, David Blanton. Their flirtatious banter had completely stopped, and Ruth became less involved in the group’s conversations.
“She was talking because she had to, not because she wanted to,” he said.
Six months later they tried again. Ruth was able to retrieve the original recordings and gave them to another company that uses AI technology. Both Ruth and David were emotional when they heard the new voice – it felt like a part of Ruth was back.
In a December interview, Ruth said, “I was amazed at how much it meant to have a voice that actually sounded like my own.
“It may sound silly, but having my own voice has given me more confidence,” she added.
Little things that I previously took for granted, like having a quiet chat together before bed or reading to my five grandchildren, suddenly took on new meaning.
Shortly after Christmas, Ruth contracted COVID. Her already limited breathing became even more difficult and she died on the morning of February 10, nine days before her 40th wedding anniversary. David held her hand all night long.
“It took us two years to say goodbye,” said David. “We agreed not to say anything.”
