Researchers have developed a brain-computer interface that can synthesize natural-sounding speech from brain activity in near real time, restoring a voice to people with severe paralysis.