Back home

Subvocal

x 0.82
Visit source

Intro

Subvocal is the product of The Subvocal Company, a San Francisco startup in Y Combinator's Fall 2026 batch. It is a pair of everyday-looking smart glasses that let you "talk" to your computer without making a sound: you silently (and almost invisibly) mouth the words, and the glasses decode them into text in real time.

Instead of cameras or microphones, Subvocal sends low-power radio waves (RF sensing) through the face and measures how they change as the speech muscles move; machine-learning models then turn those signals straight into text. The founders — Georgia Tech students Aarush Yadav and Arnav Kumar — built the first prototype at Stanford TreeHacks in February 2026. On October 30, 2026 they launched publicly on Launch YC, announced $1.3M in pre-seed funding (Afore Capital, YC, angels) and opened $199 refundable reservations, and the demo quickly became an X trending story.

Explain

Key facts

  • Company: The Subvocal Company (YC F26), San Francisco
  • Founders: Aarush Yadav (CEO, sensing hardware & firmware) and Arnav Kumar
  • Funding: ~$1.3M pre-seed — Afore Capital (led, $500k in April 2026), Y Combinator ($500k), angels
  • Interface: silent speech → text, claimed 200+ words per minute on open-ended speech with a reasonably low error rate
  • Sensing: non-contact RF sensing from the glasses frame; no cameras, no lip-reading video, no audible voice
  • Pricing: $199 refundable deposit; Early Bird price $499 (list $699); prescription lenses optional
  • Shipping: first limited batch estimated Q1 2027, US only, subject to FCC approval

How it works

  1. The glasses emit low-power radio waves through the lower face.
  2. As you silently articulate words, tiny muscle movements change how the RF signal propagates.
  3. A neural decoder maps the signal sequence directly to text, which is typed into whatever app is focused — messages, notes, code, or AI prompts.

The team plans to collect thousands of hours of population-scale RF data to pretrain a model that works on anyone after a very short calibration.

Why it matters

Voice input is fast but awkward in public; typing is private but slow. Subvocal aims to combine the speed of speech with the privacy of typing, positioning silent speech as a new primary interface for computers and AI agents. YC partner David Lieb called the live demo one of the most impressive he has seen. Caveats: the shipping hardware is still being miniaturised, and real-world accuracy across users is not yet proven.

Usage

How to get it

  1. Visit subvocal.company (also subvocal.co) and open the Preorder page.
  2. Pay the $199 refundable deposit to reserve a pair from the first production batch (US shipping only for now).
  3. Before final payment you will choose fit and lens options (prescription lenses available for an extra fee); the remaining $300 of the $499 Early Bird price is due later.
  4. Expected shipping: Q1 2027.

Good fits

  • Heavy voice-dictation users who want to dictate in offices, cafés or open-plan spaces
  • Writing prompts and messages to AI assistants hands-free and silently
  • Accessibility scenarios where speaking aloud is difficult

Tips

  • Treat it as an early-stage hardware product: specs and timing may change.
  • Watch the Launch YC post and the founders' X accounts (@AarushYadav47) for demo videos and updates.