Brain2Qwerty is a new non-invasive brain-computer interface that translates brain waves into text, offering communication without surgery.
SynthePulse Insight · AI deep readingMembers
Version 1 · 1 source
Meta releases Brain2Qwerty v2, which decodes sentences in real time from non-invasive magnetoencephalography signals using end-to-end deep learning, achieving 61% word accuracy and 78% for the best participant, with open-sourced code and data.
On June 29, 2026, Meta released Brain2Qwerty v2, a major advance in its non-invasive brain-computer interface research. The v1, released last year, already demonstrated that brain activity could be decoded without surgical implants, but v2 achieves real-time sentence decoding with accuracy approaching levels previously only possible with surgical techniques.
Invasive methods such as stereoelectroencephalography and electrocorticography have shown that neuroprosthetics combined with AI decoders can restore communication, but they are difficult to scale. Meta's non-invasive approach aims to bridge this gap, offering a new pathway for the millions of people who cannot communicate due to brain injury.
This report is based on Meta's official blog, a first-party source. All performance data (61%, 78%, 8%) are Meta's claims and have not been independently verified. Details such as training data size and participant numbers also come from Meta officially.
Primary report
Primary source