Show HN: Subvocal – Talk to your computer without making noise
Points and comments are a snapshot, not live.
Subvocal builds smart glasses that decode silent mouth movements to text using radio waves.
The authors, Georgia Tech students, built a silent speech interface using RF sensors embedded in glasses frames. The system emits radio waves that pass through the face and are modulated by tongue, lips, and soft palate movements during silent mouthing. Foundation models decode these perturbations to text. The team switched from EMG due to sweat, interference, and placement issues. They pretrained models on a large dataset collected from paid volunteers. Preorders opened for 500 units at a $200 discount, shipping Q1 2027.
What commenters are saying
Several commenters found the form factor awkward, comparing it unfavorably to throat mics or earbuds. One commenter referenced open-source alternatives like Subvox and LipFlow. A discussion emerged about lipreading's information-theoretic limits: only 12-15 visemes for 40 phonemes, making product-grade accuracy impossible. Another thread debated whether the problem of typing speed is better solved by keyboard training or ergonomic fixes. A commenter noted the landing page lacks clear explanation of how the glasses work.