Using 253 brain electrodes, researchers have restored the speech of a 47 year old woman who lost the capacity to talk 18 years ago in a stroke, by digitally synthesising her voice from prior recordings and decoding her neural activity.
The system uses machine learning to convert the mouthing of phrases into text and then speech via a replication of her own voice and enables near real time communication with only a 12% error rate for basic phrases.
From the project “BRAVO” (Brain-Computer Interface Restoration of Arm and Voice) which aims to develop neural interfaces for restoring motor function and communication.
https://www.nature.com/articles/s41593-025-01905-6