Brain2Qwerty v2
Brain2Qwerty v2 is an open-source model from Meta (Facebook Research) that decodes brain activity (ECoG/fMRI) into typed sentences, achieving state-of-the-art results on sentence-reconstruction benchmarks.
Background
This is the GitHub repository for Brain2Qwerty v2, an open-source research project from Meta (Facebook's parent company, led by AI chief Yann LeCun and CEO Mark Zuckerberg). The project is part of Meta's broader push into "brain-computer interfaces" (BCIs) — systems that decode neural activity into text or commands without requiring surgery or invasive implants. Brain2Qwerty uses non-invasive magnetoencephalography (MEG) or electroencephalography (EEG) to predict what a person is typing on a standard keyboard. The "v2" version improves on Meta's earlier model, achieving higher accuracy of decoding brain signals into typed characters in real-time. This matters because it points toward a future where people could type using only their thoughts — a paradigm shift for communication, accessibility for paralyzed individuals, and human-computer interaction in general. Current alternatives like Neuralink (Elon Musk) require surgical implants; Meta's approach aims to work with external sensors. However, the technology remains slow and limited to lab conditions; widespread consumer use is years away.