# Subvocal wants $199 down for glasses that turn silently mouthed words into text. What's proven so far

> The YC-backed startup has opened US preorders for $499 "silent speech" glasses that use radio sensing instead of cameras, and its Show HN post went viral. But there's no published accuracy figure, no final hardware, and shipping hinges on FCC approval.

Published Nov 4, 2026 by TechMeasured Staff · Gadgets · Canonical: https://techmeasured.com/gadgets/subvocal-silent-speech-glasses-499-preorder-what-is-proven

**The Subvocal Company**, a three-person San Francisco startup backed by Y Combinator, is taking US preorders for glasses that it says turn deliberately mouthed — but silent — words into text, for writing, messaging and talking to AI assistants. Reservations opened around October 31, according to EconoTimes, and the company's Show HN post on Hacker News passed 800 points this week. The pitch, from its YC profile: "Talk to your computer without talking."

Why it matters: voice is the fastest way to talk to an AI agent, but nobody wants to dictate in an open office or on a train. A silent input that works without a camera would be a genuinely new interface. The catch is that every number about how well it works so far comes from the company itself.

## How it's supposed to work

According to IBTimes and eesel, the glasses send low-power radio signals through the face and read how those signals change as the lips, tongue and soft palate move while you mouth words. AI models then decode the changes into text. There's no camera and no lip-reading from video, and the company says it only transcribes words you intentionally articulate, not a continuous recording.

Subvocal hasn't disclosed its sensing architecture, citing competitive reasons, per EconoTimes. eesel notes that a company job post lists RF sensing, EMG, EEG and mmWave as the design space. The company says it is working toward a version that works across people after a brief per-user calibration.

## The deal

| Item | What's been reported |
|---|---|
| Early-bird price | **$499** (regular price shown as $699) |
| To reserve | $199 refundable deposit; $300 balance due later |
| Not included | Tax, shipping, prescription lenses (optional paid upgrade) |
| Who can order | US delivery addresses only; buyers 18+ (eesel) |
| Estimated shipping | First shipments **Q1 2027**, "not a guaranteed delivery date" and subject to FCC approval |
| Funding | $1.3 million from Y Combinator and Afore Capital, per the company |

## What is and isn't proven

**Claimed:** prototypes decoding continuous silent speech at more than 200 words per minute, with what the company calls high or "reasonably low" error rates.

**Not published:** an accuracy or error-rate number, battery life, weight, final specs, warranty and return terms, according to eesel. IBTimes says neither speed nor accuracy has been independently verified, and that the current portable prototype is still bigger than the planned glasses. eesel also reports that the product is English-only and needs an internet connection, because decoding runs in the cloud — worth knowing for anything privacy-sensitive.

The company's own hiring post, per IBTimes, concedes the hard part: the system has to work consistently across days, after the glasses are adjusted, and for a first-time user.

The product has also moved quickly. A Georgia Tech honors-program profile from April described Subvocal as a behind-the-ear wearable reading bio-signals, with a letter-level prototype spelling out words at 30 words per minute. The founders, CEO Aarush Yadav and CTO Arnav Kumar, say they started from a hackathon prototype only months before their YC launch.

## Should you put down $199?

The deposit is refundable, so the financial risk is mostly the time your money is tied up. But you'd be reserving a development product, not buying a gadget: the hardware isn't final, there's no published accuracy, and the Q1 2027 date depends on an FCC filing that hasn't been reported as granted.

**What to watch:** an independent hands-on demo with an error rate measured on unfamiliar users, the FCC authorization, and whether Subvocal says what happens to your mouthed text in its cloud. Until then, treat 200 words per minute as a target, not a spec.


## Sources

1. [Y Combinator — The Subvocal Company](https://www.ycombinator.com/companies/the-subvocal-company)
2. [International Business Times — Could People Talk To AI Without Saying A Word? A Startup Is Making Glasses That Read Silent Speech](https://www.ibtimes.com/subvocal-glasses-silent-speech-text-3809345)
3. [eesel — Subvocal glasses: silent speech to text, price and what's proven](https://www.eesel.ai/blog/subvocal)
4. [EconoTimes — The Subvocal Company Opens U.S. Reservations for $499 Glasses](https://econotimes.com/The-Subvocal-Company-Opens-U-S-Reservations-for-499-Glasses-1767666)
5. [Georgia Tech Honors Program — Aarush Yadav and Subvocal](https://hg.gatech.edu/node/690065)