halcy 
New paper by Littlejohn etal about translating brain activity directly to speech
mimed, not fully imagined, but still - fully silent, and quality is pretty great all things considered
hot enough off the press that the DOI is apparently not properly assigned yet lol
A streaming brain-to-voice neuroprosthesis to restore naturalistic communication - Nature Neuroscience
Naturalistic communication is an aim for neuroprostheses. Here the authors present a neuroprosthesis that restores the voice of a paralyzed person simultaneously with their speaking attempts, enabling naturalistic communication.
https://www.youtube.com/watch?v=MGSoKGGbbXk man really though, we can do this with mimed speech based off of just brain signals, we can do relatively worse but still *workable* with *imagined speech*, based off of very little data (because invasive techniques), using big models *does* seem to help
I am relatively confident that if either noninvasive sensing gets good, or invasive long term sensing gets workable, speech neuroprostheses are going to be workable
A streaming brain-to-voice neuroprosthesis to restore naturalistic communication
nb: content warning for images of open skulls showing brain tissue. clinical and not like, violent, the people in question are doing perfectly fine, but they do have their skull open for surgery.
https://www.youtube.com/watch?v=XV00oTRSl1Q recent talk by Christian Herff on the topic if you want a good idea of Where We Are with speech neuroprostheses / decoding speech information directly from the brain
