Show HN: Talk to your computer without talking out loud

Aarush and Arnav of The Subvocal Company say they are developing smart glasses that turn silently mouthed words into text, aiming to make voice-dictation tools usable in libraries, offices, and public spaces. They began with facial EMG electrodes, inspired by MIT’s AlterEgo, but say sweat, electrical interference, inconsistent placement, and the discomfort of wearing electrodes made that approach impractical. They then switched to radio-frequency sensing: waves passing through the face are altered by movements of the tongue, lips, and palate, and models interpret those changes as text.

The founders describe the main technical divide as choosing between a promising sensing method and one that can work reliably as a product. Early antenna setups taped to the face or mounted on a head rig helped establish a signal, but were not suitable for glasses; they say embedding antennas in the frames required extensive engineering. They also report that differences in facial anatomy made it difficult to generalize across users, leading them to collect more training data rather than rely on algorithmic fixes. They present the current prototype’s short calibration and cross-user results as encouraging, while acknowledging continued work on edge cases and product refinement. Their plans include shipping in early 2027 and taking discounted preorders for an initial batch.