A streaming brain-to-voice neuroprosthesis to restore naturalistic communication

Littlejohn, K. T. et al. Nature Neuroscience (2025).

High-density cortical recordings and streaming neural decoding produced personalized synthesized speech in 80-millisecond increments for a participant with severe paralysis.

REPORTED FINDINGS

01The system continuously synthesized intelligible speech with reduced conversational delay.

02Models included implicit speech detection for uninterrupted operation.

03The approach generalized across several silent-speech input modalities in offline analyses.

LIMITS / CAUTION

Single-participant clinical research, implanted hardware, restricted data access, and intensive calibration limit generalization.