A wake word that never leaves the room
It is a Tuesday in 2026. A pair of earbuds sits on a kitchen counter in Yokohama. Someone says a six-syllable phrase to start a call. The earbuds wake, recognize, route the audio - and nothing crosses the internet. Half a planet away, a Honda climbs a Bay Area on-ramp; the car hears a faint siren three blocks back and softens its own audio before the driver registers it. A toy elephant in a Sao Paulo nursery answers a question in Portuguese. A Tokyo banker unlocks her phone with a glance and logs in to Mizuho without a password.
None of these devices share a brand. None of them share a country. They share a piece of software so small it can run on a microcontroller with kilobytes of memory. That software comes from a 49-person company in Santa Clara that has been doing exactly this since Bill Clinton's first term.
Sensory, Inc. is the quiet engine. The company that ships voice AI inside other companies' boxes. Three billion of them, and counting.
Three decades, small footprint, big reach
What's actually in the box
Sensory does not sell a gadget you can buy on Amazon. It sells the listening, looking and authenticating that lives inside other people's gadgets. The catalog is short, opinionated, and named with refreshing literalness.
TrulyHandsfree
The flagship. A low-power wake-word and small-vocabulary command engine that fits on a microcontroller and now powers earbuds, hearables and wearables.
TrulyNatural
Large-vocabulary speech-to-text and natural language understanding that runs entirely on-device - no cloud round trip required.
TrulySecure
Face and voice biometric verification with anti-spoofing. Picked up by Fujitsu and Mizuho Bank in late 2025.
SoundID
Recognizes sirens, glass breaks, baby cries, snores and other acoustic events that are not human speech.
VoiceHub
A self-service web tool that lets developers build a custom wake word or small command set in a browser.
Sensory Micro
The lowest-power flavor of the stack, tuned for tiny NPUs like Qualcomm Snapdragon Wear Elite and CEVA NeuPro-Nano.
Sensory has pioneered bringing speech recognition into consumer products over the past 20 years. - Todd Mozer, Chairman & CEO
A family business in microphones
Sensory was founded in 1994 by Forrest Mozer and his sons Todd and Michael, under the original name Sensory Circuits. The Mozers were not strangers to silicon - they had already co-founded ESS Technology, a chip company that went public. They were stranger to embedded speech recognition, mostly because nobody else was doing it either.
The bet: that interaction with consumer electronics should look more like interaction with people. The constraint: it had to run on the chips of the day. The result, in the mid-1990s, was the world's first commercially successful speech recognition chip - speaker-independent, 8-bit, 64KB. Suddenly a toy could understand a child. A microwave could understand a busy parent. A car key fob could understand a tired driver.
In 1999 Sensory acquired Fluent Speech Technologies, a spin-out from the Oregon Graduate Institute, picking up a deeper bench of speech science. It then quietly compounded for a quarter century - independent, profitable, with only $400,000 of disclosed outside funding to its name.
The customer list reads like CES
If you can name a major consumer electronics brand, there is a non-trivial chance you have heard yourself talk to Sensory without knowing it. Hundreds of OEMs have licensed the stack.
On-device versus over-the-wire
A rough sketch of the trade-offs Sensory has been arguing for since the cloud was a meteorology term. (Approximate, illustrative.)
Thirty years on the edge
The use cases are everywhere
Build a private voice product
Ship a hearable, doorbell or appliance that wakes to a custom phrase and recognizes commands without ever opening a network socket.
Replace a password with a face
Use TrulySecure to add face or voice verification to a phone, kiosk or banking app - with anti-spoofing baked in.
Hear what isn't speech
SoundID can flag emergency vehicles for a car, glass breaks for a security system, or a baby's cry for a monitor.
Prototype in a browser
Developers spin up a custom wake word in VoiceHub - no PhD in DSP required - and download a runnable model.
Interviews & demos
Back to the kitchen in Yokohama
The earbuds on the counter never knew the company that wrote their listening code. Neither did the Honda, the toy elephant, the Tokyo banker. That is the trick of Sensory - it is a brand whose entire identity is to disappear inside someone else's.
For thirty years the question in voice has been: who gets the audio? Sensory's answer has always been the same. The device gets it. The device keeps it. The room stays a room. In an era when "AI" usually means "send your data to a data center," that answer is not nostalgic. It is increasingly the point.
The wake word still never leaves the room. Three billion times over, and counting.