Talking to your computer is great until you try it in an open office and everyone learns what you asked ChatGPT about your rash. The Subvocal Company has a fix: glasses that let you “talk without talking.” You silently mouth words, and the glasses turn them into text, at what the company claims is more than 200 words per minute. Co-founder and CEO Aarush Yadav announced it on X on Oct. 30, along with a $1.3 million raise from Y Combinator and Afore Capital. That post has been viewed close to 12 million times and turned into an X trending story.
How do glasses hear you without sound?
They don’t hear anything, which is the point. According to the company’s Launch YC post, Subvocal uses RF sensing: the frames send low-power radio waves through your face, and AI models decode the tiny changes caused by your speech muscles as you mouth words. Digital Trends describes it as radar signals bouncing back. There is no camera, so it is not lip reading, and Subvocal says it is not always listening and cannot read your thoughts. You have to deliberately mouth each word. Phew.
The idea is to make it a keyboard replacement for your computer. The company’s site pitches it as an input method that works across apps like ChatGPT, Claude, Cursor and Slack, after a short calibration. Digital Trends also reports the company’s claim that two wearers could have a private conversation across a crowded room, which is either the future of meetings or the ultimate tool for passing notes in class.
Who is behind it?
Subvocal is part of Y Combinator’s Fall 2026 batch. Yadav and co-founder and CTO Arnav Kumar are Georgia Tech computer science students, and per their Launch YC post, the project started as a hackathon build about four months ago. That is fast. It is also why you should read the next part carefully.
What it costs, and the fine print
Preorders are open in the US only. Here is the deal, per the company’s preorder terms:
- Early bird price: $499, down from a listed $699.
- Deposit: $199 now, refundable, with the remaining $300 due later.
- Extras: taxes, shipping and prescription lenses cost more.
- Shipping: estimated for Q1 2027, but not guaranteed, and delivery depends on FCC approval.
The demo video is labeled “Prototype design,” and the Launch YC post says the current hardware is still a semi-portable board while the team shrinks it into glasses. The sleek frames in the product photos show where they are headed, not what exists today. As Digital Trends points out, the 200+ words per minute figure has not been independently verified.
The AirPods plot twist
The internet did what the internet does. Hours after the launch, one dev claimed he could do it with earbuds. Joe Blau posted on X, “I recreated subvocal in 9 minutes with one prompt for AirPods Pro 3,” and that post has topped a million views. His open-source Subvox project is a macOS prototype that reads the AirPods’ head-motion sensors and matches the movements to a handful of words you train it on. To his credit, the readme is upfront that it has no language model and picks up jaw and head movement, not your speech muscles. So it is a fun experiment, not a $499 product. But it does show how many people want this.
GadgetKing take
Silent typing at the speed of speech would be a genuine superpower, especially for anyone who lives in AI chat windows. But right now, Subvocal is a very cool demo, a refundable deposit and a lot of promises. If you are curious, the refundable deposit is a low-risk way to get in line. Just do not cancel your keyboard yet.
Sources: The Subvocal Company, Subvocal preorder terms, Launch YC, Aarush Yadav on X, Digital Trends, Joe Blau on X and Subvox on GitHub. Images: The Subvocal Company (product photos, 01 and 05) and The Subvocal Company launch video (stills, 02 to 04).
More weird wearables: the Augmental MouthPad tongue mouse, the Amazon Bee Frontier AI wearable, the Memoket Gem AI note taker and the VITURE Beast XR glasses.
