AI glasses that can read your facial muscles: How Subvocal's technology turns unspoken words into text

Developed by San Francisco-based technology startup Subvocal, the glasses reportedly use sensors and artificial intelligence (AI) to detect tiny facial muscle movements when a person mouths words without speaking aloud.

AI-powered glasses that can convert silently mouthed words into text at speeds of up to 200 words per minute could offer an alternative to conventional keyboard typing and voice-to-text software.

Developed by San Francisco-based technology startup Subvocal, the glasses, according to a report by the New York Post, use sensors and artificial intelligence (AI) to detect tiny facial muscle movements when a person mouths words without speaking aloud.

The company, which launched the product last week, as per the news outlet's report said the technology can translate these movements into text almost instantaneously, allowing users to compose emails, dictate messages and interact with AI tools without making a sound.

The news outlet quoting Aarush Yadav, who founded Subvocal in February 2026 with Georgia Tech student Arnav Kumar, said the glasses use low-power radio waves to detect movements in facial muscles.

“When you’re mouthing words, even if you’re really discreet with it, the muscles in your face and inside your mouth are moving right,” Yadav was quoted as saying. He explained that the glasses emit radio waves at frequencies similar to Bluetooth, but at significantly lower power.

As these waves interact with facial muscles, the resulting changes in signals are analysed by AI to decode the words being mouthed into text.

Subvocal glasses target office workers, AI users

The technology could offer an alternative to traditional typing and voice-based dictation, particularly in workplaces, libraries and other public spaces where speaking aloud may be inconvenient or disruptive.

“Voiceover typing is obvious,” Yadav said. “It’s extremely fast and makes you so much more productive. But when out in public in the office or in the library, we couldn’t use voice AI.”