Breaking
AirCaps joins Y Combinator's Fall 2025 batch 97% caption accuracy at ~300ms latency The New Yorker's review: "Holy shit." 16,500+ hours of real conversation processed 91% day-30 retention WIRED: "works almost eerily well" Founders built Google Glass apps at age 13 AirCaps joins Y Combinator's Fall 2025 batch 97% caption accuracy at ~300ms latency The New Yorker's review: "Holy shit." 16,500+ hours of real conversation processed 91% day-30 retention WIRED: "works almost eerily well" Founders built Google Glass apps at age 13
Company Profile · Hardware / AI

AirCaps Is Putting Subtitles on Real Life

Two founders who have been obsessed with smart glasses since they were teenagers built a 49-gram pair that captions any conversation, translates 60-plus languages, and remembers the meeting for you.

There is an uncomfortable statistic that AirCaps likes to repeat: most people retain only about half of an in-person conversation, and forget roughly 70 percent of it within two days. For someone with hearing loss, the number is worse - not forgotten later, but missed in the moment. AirCaps built a pair of glasses around that gap. Put them on, and the words other people say appear as a thin band of text along the bottom of your lenses, in near real time, whether you are at a dinner table, a supermarket checkout, or a sales call.

The company, which goes by AirCaps and was previously known as TranscribeGlass, is part of Y Combinator's Fall 2025 batch. Its pitch is deceptively narrow. Not a general-purpose computer for your face, not a heads-up display for notifications - just captions, translation, and a memory of what was said. That focus is the whole idea.

49gFrame weight
97%Caption accuracy
~300msLatency
60+Languages

01 / What it doesReading the room, literally

AirCaps are lightweight augmented-reality glasses with a private binocular display. A four-microphone beamforming array listens to the room and isolates the person in front of you from the noise around them. Speech is turned into text in the cloud and pushed back to the lenses fast enough - the company cites roughly 300 milliseconds - that the caption keeps pace with the talker. The display leaks less than two percent of its light, so the person across the table cannot see that you are reading them.

On top of captions sit two more layers. Translation covers 60-plus languages with context-aware handling of idioms, so a conversation in another language shows up in yours. And for professionals, an "AI meeting intelligence" layer takes notes, pulls out action items, labels who said what - up to 15 distinct speakers - and keeps a searchable history you can go back to later.

On-lens caption accuracy (company-reported)
Quiet room97%
Busy cafe93%
Free tier90%+
Figures as stated by AirCaps. Accuracy in real-world noise is the number the company stakes its reputation on - the four-mic array is tuned for restaurants and parties, not silent labs.

02 / Who it's forStart with the people who need it most

AirCaps began with the Deaf and Hard of Hearing community, and the testimonials read less like product reviews than like small returns to normal life.

"What I realize is that this has the potential to change my life. I could start going to the theatre again." Susan Goldhor, Cambridge, MA

From there the audience widened without the product changing much. International travelers use it as a live translator that does not require holding up a phone. Language learners use it as a running answer key. And professionals - sales reps, doctors, founders, executives, lawyers - use the meeting layer to stop taking notes and start paying attention. The American Bar Association even ran a product note on the glasses in early 2026. It is the classic accessibility-first pattern: build for the hardest case, and the easier cases come along for free.

"I was blown away that I understood everything. This is by far the most amazing thing I've ever experienced in my life." Leah Ewart, AirCaps user

03 / The problemEvery existing fix breaks eye contact

The reason this is hard, and the reason nobody had nailed it, is that the alternatives all cost you something social. A phone held up mid-conversation kills eye contact and signals distraction. Earbuds and hearing aids amplify sound - including the noise you were trying to escape - without ever showing you the words. A laptop open on the table changes the whole tenor of a dinner. AirCaps' bet is that the display belongs where your attention already is: on the face of the person talking.

How a sentence becomes a caption
STEP 1

Capture

4-mic beamforming array isolates the speaker from ambient noise.

STEP 2

Transcribe

Audio streams to the cloud; speech becomes text and is diarized by speaker.

STEP 3

Translate

Optional pass converts 60+ languages, keeping idioms intact.

STEP 4

Display

Text lands on the lens in ~300ms, under 2% light leakage.

The pipeline behind the glass. A phone provides the Bluetooth and internet link; the glasses handle capture and display.

04 / The foundersEleven years of one obsession

The company was founded in 2024 by Madhav Lavakare and Nirbhay Narang, though the origin story starts much earlier. Lavakare, now CEO, was writing apps for Google Glass and building voice assistants on Raspberry Pi at age 13. He studied computer science at Yale and researched audio AI at the MIT Media Lab, where his thesis went straight at the core problem: pulling a clean transcript out of noisy, multi-speaker audio. Narang, the CTO, graduated summa cum laude in computer science from Cornell, built voice AI for AR glasses in high school, and was an early engineer at multiple YC startups working on conversational AI.

Between them, more than a decade spent on the unfashionable belief that smart glasses were worth building - long after the rest of the industry had written Google Glass off.

05 / The businessSell the hardware, keep the software honest

The glasses start at $599 and are HSA/FSA eligible. Where AirCaps departs from the usual subscription trap is the free tier: captions in nine languages, unlimited, forever, with five hours a month of the paid features thrown in. AirCaps Pro, at $20 a month, unlocks the 60-plus languages, higher accuracy, and the meeting intelligence. Prescription lens holders run $39, with support across a wide -16 to +16 diopter range. You can buy the device and never pay another cent, which is a deliberate signal to a community that has been sold expensive, subscription-locked assistive tech before.

$599Device price
$20/moPro tier
8 hrsBattery life
15Speakers labeled

06 / TractionThe metric that can't be faked

Hardware companies love to quote units shipped. AirCaps prefers retention. The company says 91 percent of users are still wearing the glasses 30 days in, that power users average six-plus hours a day, and that it has processed more than 16,500 hours of real conversation. On the revenue side it reported $93,000 in a single month, growing 6.5x month over month. An earlier version of the product, under the TranscribeGlass name, pulled 75 million-plus TikTok views and a following in the six figures.

Reported month-over-month revenue growth
Month 11x
Month 2~4x
Month 36.5x
Curve steepening. A reported $93K month capped the run - impressive for hardware, though early-stage figures are self-reported and worth reading as a trajectory, not a guarantee.

07 / The fieldA narrow product in a crowded race

Smart glasses are having a moment. Meta's Ray-Ban line, Even Realities, Brilliant Labs, and Vuzix-based captioning systems are all chasing some version of the computer-on-your-face future. AirCaps is deliberately not doing that. It is not trying to show you notifications, take photos, or run an assistant that does everything. It does one job - understanding conversation - and it competes as much with hearing aids and captioning apps as it does with other glasses. In a space where most players are widening, AirCaps narrowed, and the retention numbers suggest the narrowing is what makes people keep the glasses on.

The reviews have not been shy. WIRED said the glasses work "almost eerily well." The New Yorker's reaction ran to two words: "Holy shit." Forbes, Fast Company, Smithsonian Magazine, and Scientific American have all covered the product. Whether AirCaps becomes the default way people caption the world or an early proof of concept for a bigger platform, it has already done the hard part - built something people wear all day and describe in the language of getting their lives back.

#smart-glasses#ar#real-time-captions #live-translation#accessibility#deaf-and-hard-of-hearing #wearables#speech-recognition#yc-f25 #meeting-intelligence#assistive-technology