A woman in her 40s who lost the use of her limbs and her ability to speak due to a stroke has regained her voice after 18 years. This was made possible by implanting a device in her brain that reads her thoughts in real time. While similar experiments have succeeded in the past, there was a delay in converting thoughts into speech. This time, it essentially implements thoughts into voice in real time, which is expected to greatly help patients with speech disabilities recover their daily lives.
A joint research team led by Edward Chang from the University of California, San Francisco (UCSF), and Gopala Anumanchipalli from the University of California, Berkeley (UC Berkeley) announced on 1st that they successfully expressed the thoughts of a woman in her 40s who was paralyzed from a stroke using a brain-computer interface (BCI). The research results were published that day in the international journal 'Nature Neuroscience.'
BCI is a technology that connects the brain to a computer to restore bodily functions or transmit human intentions to external devices. Electrodes implanted in the brain of paralyzed patients capture thoughts and manipulate robots or computers accordingly, implementing them in text or speech.
The clinical trial involved a 47-year-old woman who had been paralyzed and lost her ability to speak due to a stroke 18 years ago. The BCI device reduced the speech conversion delay to a level of several milliseconds (1 millisecond is one-millionth of a second), achieving real-time voice implementation.
Previously, the research team developed a device that reads brain signals in 2023, but there was a delay of several seconds between thought and voice conversion, which limited practical applications. In real conversations, even a delay of one second makes communication difficult. At that time, they opted to implement speech along with a digital avatar to enhance communication.
The research team presented sentences to the woman and asked her to think about speaking, then measured the brain signals. Approximately 1,024 words were included in the data training sentences. The BCI device captured signals from the motor cortex, the brain area responsible for language ability, and converted the woman's thoughts into speech.
The voice synthesized by the computer was recreated based on recordings made before the woman had her stroke. As a result, the woman succeeded in thinking a sentence and reading it in voice in 80 milliseconds. Words not used in training were also converted to speech normally.
The research team noted, "The real-time voice generation system has the potential to maintain the flow of natural conversation," adding, "We expect this research will help restore communication abilities for patients who have lost their language skills."
However, because this study involved only one participant, further research is needed to confirm the performance and side effects of the BCI device. The research team said, "Additional studies involving more patients are necessary," while adding, "If patients who have lost their language skills can communicate naturally and smoothly, their quality of life will increase significantly."
References
Nature Neuroscience (2025), DOI: https://doi.org/10.1038/s41593-025-01905-6
Nature (2023). DOI: https://doi.org/10.1038/s41586-023-06443-4