An era where you can enter text silently just by moving your mouth has arrived. US startup Subvocal has revealed typing glasses powered by muscle recognition capable of input speeds exceeding 200 words per minute.
On October 30 local time, Subvocal announced the opening of pre-orders alongside news that it secured $1.3 million in funding from Y Combinator and Afore Capital. The early bird price is set at $499 (approx. 670,000 won), with the product page also listing a regular price of $699 (approx. 940,000 won).
Subvocal is a system that reads the movement of speech muscles using radar radio waves rather than cameras. Because radio waves propagate through the air, this method can detect movement beneath the skin without direct contact. Subvocal also clarified that it only recognizes words formed by intentionally moving the mouth and does not read thoughts. In demonstration videos, the device showed rapid typing achieved merely by subtly mouthing words.
The project originated at Stanford University's hackathon, TreeHacks. In a post on X, Subvocal shared that they initially attached electromyography (EMG) sensors to the face. Using data from silently pronouncing the 26 letters of the alphabet, they trained a model within 24 hours, achieving a character classification accuracy of around 80%.
Since then, they tested four methods: EMG, video, inertial measurement units (IMUs), and radio waves. EMG requires skin contact, and its accuracy fluctuates significantly with each use. Webcam video brought the word error rate down to 17%, but because it only tracks the lips, it struggles to distinguish sounds with similar lip shapes, such as P and B. IMUs only capture overall facial movements, revealing clear limitations as the vocabulary expands. In contrast, Subvocal claims its current prototype glasses can read continuous speech after about a minute of calibration even for first-time users, with a word accuracy of 85–92%.
The company cited a 2022 paper on radar-based silent speech recognition published in the journal Scientific Reports as the foundation for its radio wave approach.
Products targeting silent speech input have appeared before. On January 29, Apple acquired Q.ai, an Israeli startup that reads silent speech through micro-movements of facial skin, for $2 billion. In September last year, AlterEgo unveiled a face-mounted wearable originating from MIT Media Lab research, demonstrating a video of typing without speaking.
The industry appears to be increasingly focusing on silent speech recognition that requires neither typing by hand nor making any sound.
First shipments of Subvocal's glass-type device are expected in Q1 2027. Subvocal added that the timeline is subject to change as the product remains under development.
