Preprint
Brain-to-Text Decoding: A Non-invasive Approach via Typing
Jarod Levy et al., arXiv (preprint), 2025
Read the originalRead on October 17, 2026
- What it is
- A Meta AI research preprint describing Brain2Qwerty, a system that decodes the text of sentences from brain recordings made with magnetoencephalography (MEG) or electroencephalography (EEG) while healthy volunteers type. The abstract reports character error rates for both recording methods.
- How much weight it carries
- A preprint that has not been confirmed as published in a peer-reviewed journal in the sources read here. It tests healthy volunteers who type, so it does not show that the method works for people who cannot type or speak. Accuracy figures are character error rates, not words per minute.
Facts taken from this source
- The Brain2Qwerty preprint tests its decoder in a group of 35 healthy volunteers.
“demonstrate its efficacy in a cohort of 35 healthy volunteers” (Abstract)
- In a 2025 preprint, the Brain2Qwerty decoder reached an average character error rate of 32 percent with magnetoencephalography (MEG) recordings.
“With MEG, Brain2Qwerty reaches, on average, a character-error-rate (CER) of 32% and substantially outperforms EEG (CER: 67%).” (Abstract)
- The same 2025 preprint reports a character error rate of 67 percent when the decoder uses electroencephalography (EEG) recordings instead of MEG.
“substantially outperforms EEG (CER: 67%)” (Abstract)