When doctors told her that her tongue and voice box would need to be removed to save her life from the cancer that had invaded her mouth, Sonya Sotinsky sat down with a microphone and prepared to record the words she would never say again.
Top phrases she wrote to her husband and two daughters included “Happy Birthday” and “I’m proud of you,” as well as “I’m perfect for you” to clients of the architecture firm she co-owns in Tucson, Arizona.
Thinking about the grandchildren she longed to have one day, she also recorded more than a dozen children’s books, from the Eloise series to Dr. Seuss, that she would one day play to them at bedtime.
But one of the biggest categories of audio files she saved was a series of expletives and profanity. If the voice is the main expression of personality, irony and profanity are essential for Sotinsky.
“It’s very, very frustrating when you can’t use your voice. Other people project your personality onto you. I silently screamed and screamed at the lack of a voice,” Sotinski said recently, referring to rudimentary voice technology and handwritten notes before stumbling upon the latest workaround. “Literally what?”
After battling invasive oral cancer at the age of 51, Sochinsky was forced to confront the significance of the human voice. She felt that her distinct intonation, intonation, and slight New Jersey accent were vestiges of her identity. And she refused to be silent.
Although doctors and insurance companies saved her life, they showed little interest in saving her voice, she said. So she set out to independently research and identify artificial intelligence companies that could do just that. It used the recordings Sotynski had saved of her natural voice to create an exact replica that is now stored in an app on her phone, allowing her to type and speak again with all the emotion and sarcasm.
“She’s got her sass back,” said Sotynski’s daughter, Ella Fuentevilla, 23. “My sister, father, and I all cried when we heard her AI voice. It’s surprising how similar they are.”
“Your voice is your identity”
It took nearly a year for doctors to discover Sochinski’s cancer. She repeatedly complained to her orthodontist and dentist about jaw pain and strange sensations under her tongue. Then, as she drank the water, water started dripping down her chin. When the pain became so severe that she was unable to speak by the end of each day, Sotinski insisted that she see an orthodontist for further evaluation.
“A shadow fell on his face, and I saw it when he leaned back. It was that look you didn’t want to see,” she said.
That’s when she started recording. In the five weeks between diagnosis and surgery to remove her tongue and entire voice box (medical terminology: glossectomy and laryngectomy), she saved as much of her voice as she could.
“Your voice is your identity,” says Sue Yom, a radiation oncologist at the University of California, San Francisco, where Sotinski was treated. “Communication is not only the way we express ourselves and relate to other people, but also the way we understand the world.”
“When you lose your voice, you can’t hear yourself thinking out loud, you can’t hear yourself interacting with other people,” Yom says. “It affects how your mind works.”
He added that people who lose their voice boxes are at higher risk of long-term psychological distress, depression and physical pain than those who retain their voice boxes after cancer treatment. Nearly a third have lost their jobs and social isolation can be severe.
Most laryngectomy patients learn to speak again using an electrolarynx, a small battery-powered box held against the throat that produces a monotonous, mechanical sound. But Sotinski knew it wouldn’t work for her without a tongue to form words.
When Sotinski underwent surgery in January 2022, AI voice was still in its infancy. The best technology she could find produced a synthetic version of her voice, but it was still flat and robotic, and people struggled to understand her.
She managed to get by until mid-2024, when she read an article about tech companies using generative AI to recreate all of our natural intonations and emotions.
Companies can now recreate a person’s voice from snippets of old home movies or one-minute voicemails, but the optimal range is 30 minutes.
Sotinski would spend hours reading children’s books aloud to save money.
“Eloise saved my voice,” Sotinski said.
Now, she types what she wants to say into Whisper, a text-to-speech app on her phone, and the AI voice is translated and broadcast through a portable speaker.
Yom said most doctors and speech-language pathologists who treat head and neck cancer patients are unaware that AI software can be used in this way, and because they are focused on saving lives, they often don’t have the bandwidth to prompt patients to record their own voices before they lose their voices during surgery.
Health insurance companies similarly prioritize treatments that extend life over those that improve quality of care, and typically avoid covering new technologies until data proves their actuarial value.
Sotinski and her daughter argued for months with insurance claims adjusters at Blue Cross Blue Shield of Arizona, but the insurance company refused to reimburse Sotinski for the $3,000 she had spent on the initial assistive speech technology.
“Apparently vocalization is not considered a medical necessity,” Sotinski quipped in his sarcastic AI voice.
Sotinski currently pays a monthly fee of $99 out of pocket for the AI voice clone.
“Medicare insurance covers both routine and life-saving care, but communication aids are typically not covered,” said Teresa Joseph, spokeswoman for Blue Cross Blue Shield of Arizona. “As AI provides opportunities to impact health, we expect applicable standards to evolve across the country.”
Research could lead to insurance coverage
Sotinski decided to use her newfound voice to help others regain their voices. She left her career in architecture to build voicebanknow.com, a website detailing her voice banking journey. She tells her story at conferences and webinars, including an oncology conference Yom hosted in Denver for 80 scientists.
She added that breast cancer survivors faced similar struggles in the 1980s and ’90s. Her insurance company initially refused to cover the cost of breast reconstruction following the mastectomy, claiming the surgery was cosmetic and unnecessary.
It took years of patient advocacy and carefully crafted data showing that Reconstruction had a significant impact on women’s physical and mental health before the federal government mandated coverage in 1998.
De los Santos and Yom said research data on AI voice cloning is likely to follow a similar path, ultimately proving that fully functional, natural voices not only lead to better lives, but also longer lives.
In recent months, Sotinski’s AI voice literally helped save her life. The cancer had returned in her lungs and liver. Thanks to her voice, she was able to communicate with her doctors and fully participate in the development of her treatment plan. It showed her how “medically necessary” it is to have a voice.
She noticed that doctors and nurses were taking her more seriously. They didn’t miss a beat, as people often did when she relied on robotic, synthetic voices. They seemed to see her as a more complete person.
“If someone can only communicate a few words at a time and cannot elaborate or coordinate more fully, it stands to reason that we cannot detect that they are thinking deeper,” she said. “Being able to interact with your care team in a more seamless way is essential.”
Sotinski, now 55, said that despite doctors’ success with her latest cancer treatment, she is facing the reality that she will likely die much sooner than she had hoped, and is confronting her odds in a new way.
She realized once again how important her voice is to maintaining her outlook on life and sense of humor in the face of death.
“I tend to forget and think I’m okay, but actually, this is going to last forever. Emotionally, I start getting cocky again. And I was like, this is it. Oh shit, we’re not playing. this cancer is real” Sotynski said with a mischievous grin as he typed his next phrase.
“Sarcasm is part of my love language.”
This article is a partnership with KQED and NPR.
