Brain2Qwerty v2 Decodes Sentences at 61% Accuracy From Non-Invasive Brain Signals
Updated
Updated · Meta AI · Jun 29
Brain2Qwerty v2 Decodes Sentences at 61% Accuracy From Non-Invasive Brain Signals
3 articles · Updated · Meta AI · Jun 29
Summary
Brain2Qwerty v2 reached 61% word accuracy in real-time sentence decoding from non-invasive brain recordings, with its best participant hitting 78% and more than half of sentences decoded with one word error or less.
22,000 sentences from nine volunteers powered the system, which used magnetoencephalography recordings during 10-hour typing sessions and end-to-end deep learning instead of hand-crafted neural-event pipelines.
The result sharply outperformed other non-invasive methods at 8% word accuracy, while researchers said performance improved log-linearly with more data, suggesting scaling alone could narrow the gap with surgical approaches.
Full training code for Brain2Qwerty v1 and v2 is being released, and partner BCBL is releasing the v1 dataset to support open neuroscience research and communication tools for people with brain lesions.
This AI reads typists' minds, but can it give a voice to those who can only imagine speaking?
If an AI can translate our brainwaves into text, who truly owns the thoughts that it decodes?
Brain2Qwerty v2: Non-Invasive BCI Reaches 22% Character Decoding Accuracy with MEG and Deep Learning
Overview
Brain2Qwerty v2, announced in March 2025, marks a major leap in non-invasive brain-computer interfaces by using a sophisticated deep learning model to decode neural signals for communication. Researchers analyzed how the system’s typing protocol triggered different brain responses, finding that Magnetoencephalography (MEG) outperformed Electroencephalography (EEG) in classifying hand movements and decoding characters. The deep learning model, with its advanced components, significantly improved decoding accuracy over traditional methods. These advances show that Brain2Qwerty v2 can more effectively translate brain activity into text, paving the way for safer and more intuitive communication technologies.