Get the newsletter
Apple · Wearable · Filed Jul 8, 2026 · USPTO published Nov 19, 2026 · This week's release

Apple Files Patent for a Wearable That Hears Words You Never Say Aloud

Apple's patent application describes an earbud or pair of glasses that reads the tiny muscle movements in your face, so you can talk to an AI assistant without making a sound. It even tries to tell whether you paused to think or finished your question.

An earpiece projects light onto facial muscles to detect silent speech and send the words directly to a smartphone. Drawing from patent filing US 2026/0369274 A1.
An earpiece projects light onto facial muscles to detect silent speech and send the words directly to a smartphone.All 14 drawings ↓
Publication number
US 2026/0369274 A1(US20260369274A1)
Applicant
APPLE INC.
Filed · Published
Jul 8, 2026 · Nov 19, 2026
Inventors
Yonatan Wexler, Asaf Kagan, Oren Amsalem, Michael Doron
Status when we published
Waiting for an examiner
Document
20 claims
US classification
704/231
Parent application
Continuation of PCT/IB2025/059025 (filed Sep 8, 2025)(PCTIB2025059025)

What Apple's silent-speech wearable actually does

Ever tried to ask your phone a question on a packed train and felt silly speaking out loud? Apple has a patent application for a head-worn gadget that lets you mouth the words with no sound at all and still be understood.

The idea is that a small device on your ear or face senses the tiny muscle movements your face makes when you form words. Software turns those movements into what you meant to say, then answers you privately, perhaps through a speaker in your ear. The filing also covers some friendly extras: a reminder of an earlier topic after you have moved on to another, and replies that copy your own speaking style.

It also tries to judge whether you stopped mid-question because you were thinking or because you were done, so an assistant does not jump in too early.

From the filing · CLAIM 1
… at least one sensor incorporated with the housing and configured to output signals indicative of communication-related neuromuscular activity of the individual …

Translation: Sensors built into the headset track the tiny muscle movements you make when trying to speak.

How the sensor reads tiny movements in your face

The claimed system has three parts: a housing worn on the head, a sensor, and a processor. The sensor produces signals tied to neuromuscular activity (the nerve and muscle activity that happens when you form words). The processor works out the substance of a conversation event, meaning what you actually said or meant, and then generates an output such as audio, text or a command.

The description fills in the how. The housing can be an earbud, glasses or a headset. In the main example, a light source shines laser light on your cheek and a detector watches how it scatters off the skin. The light forms a grainy pattern (called speckle) that shifts when facial muscles move, even by tiny fractions of a millimeter. A trained neural network (software that learns patterns from many examples) turns those shifts into words, and the filing says it can be fine-tuned to a new wearer within minutes. One claim also covers sensing activity tied to a cranial nerve (the nerves that run from the brain to the face).

The dependent claims add specific behaviors:

  • Topic reminders: after you move to a second topic, you get a private reminder of the first.
  • Speaking style: it picks up your cadence, tone and volume and answers in a matching manner.
  • Pause checks: it decides whether a pause means you are still forming a question or are done, then holds the query or sends it to the assistant.
  • Turn-taking: it tracks the back-and-forth with another person or a large language model (the AI behind chatbots).
From the filing · THE ABSTRACT
… analyze the received signals to determine substance of at least one conversation event associated with the individual …

Translation: A computer processor decodes those muscle signals to figure out what you are silently trying to say.

What silent speech could change for Apple's assistant

If you have ever hesitated to talk to your phone on a train, in an office or in a waiting room, this is the problem being attacked. A device that reads your mouthed words lets you ask an assistant a question or send a command privately, with no sound for anyone nearby to hear. The filing also shows two people each wearing a device and talking silently to one another.

For Apple, the pieces line up with hardware people already wear: earbuds and headsets are among the head-worn housings the filing names, and it lists glasses and extended reality headsets too. The assistant-focused claims (reminders of earlier topics, waiting out your pauses, matching your speaking style) aim to make a voice helper feel more like a conversation than a string of commands.

This is the 14th Apple filing we've tracked since May on our wearables reading your body watch, following one on head-tilt voice cutoff and another on gesture voice shortcuts.

Source: US 2026/0369274 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Apple patent news breaks.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

The drawings · 14 sheets

14 drawing sheets from US 2026/0369274 A1 · click any drawing to enlarge

An earbud uses light to read unspoken words, drawing sheet 1 from US 2026/0369274 A1
Sheet 1 An earbud uses light to read unspoken words
Smart glasses track tiny cheek movements to detect silent speech, drawing sheet 2 from US 2026/0369274 A1
Sheet 2 Smart glasses track tiny cheek movements to detect silent speech
Pairing a headset with an ear sensor reads silent words, drawing sheet 3 from US 2026/0369274 A1
Sheet 3 Pairing a headset with an ear sensor reads silent words
Your phone tracks facial light reflections to catch silent words, drawing sheet 4 from US 2026/0369274 A1
Sheet 4 Your phone tracks facial light reflections to catch silent words
Hardware parts that send your face reflections to the cloud, drawing sheet 5 from US 2026/0369274 A1
Sheet 5 Hardware parts that send your face reflections to the cloud
Sensing tiny skin shifts before and after facial muscles move, drawing sheet 6 from US 2026/0369274 A1
Sheet 6 Sensing tiny skin shifts before and after facial muscles move
Light speckle patterns change when hidden speech muscles flex, drawing sheet 7 from US 2026/0369274 A1
Sheet 7 Light speckle patterns change when hidden speech muscles flex
Software tasks that decode silent speech patterns into words, drawing sheet 8 from US 2026/0369274 A1
Sheet 8 Software tasks that decode silent speech patterns into words
Show the other 6 sheets
Matching cheek movements to specific unspoken words or expressions, drawing sheet 9 from US 2026/0369274 A1
Sheet 9 Matching cheek movements to specific unspoken words or expressions
Shining a grid of light points across key facial muscles, drawing sheet 10 from US 2026/0369274 A1
Sheet 10 Shining a grid of light points across key facial muscles
Asking a phone for financial help without speaking out loud, drawing sheet 11 from US 2026/0369274 A1
Sheet 11 Asking a phone for financial help without speaking out loud
Correcting a spoken voice assistant with a silent reply, drawing sheet 12 from US 2026/0369274 A1
Sheet 12 Correcting a spoken voice assistant with a silent reply
Two people holding a silent conversation using wearable sensors, drawing sheet 13 from US 2026/0369274 A1
Sheet 13 Two people holding a silent conversation using wearable sensors
Steps for turning tiny muscle twitches into silent digital words, drawing sheet 14 from US 2026/0369274 A1
Sheet 14 Steps for turning tiny muscle twitches into silent digital words
Patent filing page

Watchlists this filing belongs to

All watchlists →
Reader comments

No comments yet

Start the discussion

Real name or a handle, either is fine. Comments are read by a person before they appear, so allow a little time. Keep it about the filing.