“A shadow cast over his face. I saw it when he leaned back — that look you don’t want to see,” Sotinsky said. “That look that I probably shouldn’t have missed this.”
Head and neck cancer rates in the U.S. have declined steadily in recent decades, in step with declines in smoking, chewing tobacco and drinking. HPV-related throat cancers, contracted through oral sex and even kissing, have gone up, accounting for 70% of roughly 60,000 new head and neck cancer diagnoses each year. Uptake of a 2006 HPV vaccine that can prevent such cancers has been modest, with about 60% of adolescents and less than 16% of adults up to date on their shots.
Sotinsky’s cancer likely derived from the third common cause: a genetic mutation or day-to-day cut, bite or burn to the mouth that stimulated abnormal cell growth during the healing process. By the time Sotinsky arrived at UCSF’s oncology department for treatment, the cancerous cells were so widespread that surgeons had to remove her entire tongue and her voice box. In medical terms, a total glossectomy and laryngectomy.
Dr. Sue Yom, a radiation oncologist, stands in a room used for radiation treatments for cancer patients at UCSF’s Mission Bay campus on May 16, 2025. (Beth LaBerge/KQED)
“Laryngectomy is one of the most terrifying and impactful outcomes for a patient surviving cancer,” said Dr. Sue Yom, Sotinsky’s radiation oncologist at UCSF.
People who lose their voice box, she added, are at higher risk for long-term emotional distress, depression and physical pain compared to those who retain it after treatment. Close to a third lose their job and social isolation is profound. Mouth and throat cancer survivors often become unable to swallow or eat solid foods around the same time they become unable to speak, making it difficult to join family gatherings that revolve around food and take place in noisy settings where assistive speaking technologies don’t work well.
“Your voice is your identity,” Yom said. “Communication is not only how we express ourselves and relate to other people, but also how we make sense of the world. When the voice is no longer available, you can’t hear yourself thinking out loud, you can’t hear yourself interacting with other people. It impacts how your mind works.”
Sotinsky understood this instinctively when she sat down to record her phrases and curse words. Thinking about the grandchildren she desperately hoped to see born one day, she also recorded herself reading more than a dozen children’s books, from Eloise to Dr. Seuss, to play for them one day at bedtime. In the five weeks between her diagnosis and surgery, she banked as much of her voice as she could manage.
Most laryngectomy patients learn to speak again with an electrolarynx, a small battery-operated box held against the throat that produces a monotonic, mechanical voice. But without a tongue to shape her words, Sotinsky knew that wouldn’t work. At the time of her surgery in January 2022, AI voices were still in their infancy. The best technology she could find yielded a synthetic version of her voice, but it was still flat and robotic and others strained to understand what she said or inferred the wrong meaning.
Sonya Sotinsky sits at Crane Cove Park in San Francisco on March 25, 2025. (Beth LaBerge/KQED)
She got by until mid-2024 when she discovered Eleven Labs, an AI company that relied on her children’s book readings and banked phrases to build the full, rounded voice clone she uses today. It gets funneled through a text-to-speech app on Sotinsky’s phone, called Whisper, that translates what she types into her AI voice and broadcasts it through portable speakers. Her doctors were awestruck when they heard it.
“Sometimes people inadvertently equate loss of voice with a loss of intelligence,” Yom said. “And what Sonya really shows is, ‘I have a lot of ideas. I have a lot to say. I can take my place with anybody speaking my mind.’”
Yom was blown away by the force and beauty of Sotinsky’s AI voice, but also all the effort and resourcefulness that went into making it happen. Most doctors and speech therapists who work with head and neck cancer patients don’t realize AI software can be used this way, and with their focus on saving lives, Yom said, they often don’t have the bandwidth to encourage patients to record their voices before losing them.
