Silent speech AI system uses ultrasound imaging of vocal tract movements to translate silent articulation into text

A post from @beffjezos is circulating a silent speech AI demo that is described as using ultrasound imaging of vocal tract movements to translate silent articulation into text, with the main claim framed around a model that generalizes to new users without user-specific calibration. The public sourcing visible here is still narrow, so the safest read is to stick to what the post says and what linked context from @vadi_ms and @alephneuro appears to show, not treat it as independently settled.