Subvocal is pitching glasses that turn silently mouthed words into text: dictation without making the whole room an audience. Its website claims a speed of 200-plus words per minute for messages, notes and conversations with AI. Source
In a November 17 post, @AarushYadav47 described a silent conversation using Subvocal: “I couldn’t hear Arnav, but the glasses could.” Source
Your mouth still has a job
Subvocal says the input requires “deliberate silent articulation”: users intentionally mouth words without speaking aloud. Someone up close may notice slight movements, according to its FAQ. Source
The company also says it uses neither cameras nor camera-based lip reading for silent-speech input. The website doesn’t explain what sensors it uses instead. Subvocal says it decodes intentionally articulated words, not arbitrary or involuntary thoughts. Merely thinking of a devastating reply won’t enter it for you. Source
Subvocal says its input doesn’t depend on hearing your voice, targeting situations where speaking aloud is awkward or background noise interferes with ordinary dictation. Its website specifically promotes directing coding agents without disturbing coworkers. Source
Prototypes and planned app support
Subvocal describes its current devices as prototypes that decode open-ended silent speech in real time. Its website acknowledges errors but supplies no numerical accuracy benchmark, independent test results or methodology for its speed claim. The FAQ also says setup will require a short personal calibration process, without specifying a duration. Source
The company says it’s building Subvocal as an input method that users “should be able to use” with tools including ChatGPT, Claude, Cursor and Slack, alongside email, documents and messaging. Source
The website invites reservations but lists no price or shipping date. Source
Comments