Brain2Qwerty: Brain-to-text decoder by Meta
Meta has developed Brain2Qwerty, a non-invasive brain-to-text decoder that reconstructs sentences from brain activity captured via MEG and EEG. The system decodes up to 80 characters per minute from participants typing sentences, achieving a mean word error rate of 32.5% for a 52-word vocabulary.
Background
Meta (Facebook's parent company) has published research on Brain2Qwerty, a system that decodes brain signals into typed text using non-invasive (MEG/EEG) brain recording. This is not a product — it's a research paper.