The technology, led by neurosurgeon Edward Chang, translates neural activity from the vocal motor cortex into text and speech. The system relies on a phenomenon known as persistent somatotopy, where cortical neurons continue to generate signals for speech even when the connections to the body are severed by injury.

The patient, who goes by the nickname Pancho, suffered a brainstem stroke in 2003 at age twenty. For years, he communicated using a pointer attached to a baseball cap, achieving a speed of only five words per minute. His implant consists of a flexible array of 128 to 200 electrodes placed on the surface of his brain, connected to a computer for analysis.

Unlike previous devices that tracked cursor movements or eye tracking to type letters individually, this system uses an artificial neural network to identify phonemes. This approach allows the model to recognize a vocabulary covering nearly 90 percent of the English language, significantly expanding the range of words a patient can produce.

The BCI also aims to restore nonverbal aspects of communication. By interpreting activity in vocal motor regions, the system can generate a humanlike avatar that conveys the patient’s personalized voice tone and facial expressions. This addresses the loss of emotional communication often experienced by individuals with paralysis.

While earlier brain-computer interfaces managed about 10 words per minute, the new system achieves an output of almost 80 words per minute. The research represents a significant step toward restoring natural speech capabilities for individuals with severe motor impairments.