Sector · Live translation

Live translation on the lens

The most sold feature of smart glasses, and the one with the widest gap between the demo and the day. Here is what people who used it in the field reported, with the address of every claim.

Translation is the reason most people try smart glasses, and it is the feature every maker puts first. It is also the one where the promise and the day pull apart, and the cases below say so in the words of the people who ran them.

What holds up is the short exchange. A stall team at the Canton Fair served buyers from Argentina, Brazil and Kazakhstan with face to face translation and had them queue to try it. A reviewer who took the same idea into streets, restaurants, a factory visit and business meetings landed on a precise verdict: use the glasses to listen, and the phone to show the other person the text. Nobody in these sources replaced a human interpreter.

What breaks is the room. Natural conversation, several speakers, a strong accent and noise all cost accuracy, and one reviewer watched a laugh in the middle of a sentence throw the translator into an unintelligible line. A live stage demo by Google stopped translating after about fifteen seconds. Language coverage is uneven too: the same reviewer found no German on the device he takes to Berlin every year.

The second audience is not the tourist

Live captioning on the lens was built for deaf and hard of hearing people before it was built for travellers, and that line of work is older and better documented. XRAI shipped captions on AR glasses in 2022 with a deaf cofounder; the MentraOS community keeps Live Captions and Live Translation as open source apps under Apache 2.0, running on more than one brand of glasses; a Texas school assessed reading and translating printed text out loud for blind and low vision students. If you are building here, that is where the demand is proven.

The wall you should design around

New York banned smart glasses with a camera or microphone from every state court in July 2026, and wrote no exception for a deaf visitor who depends on captions on the lens. A California bill would make wearable recording a criminal misdemeanour, including the act of switching the recording light off. Any product in this sector has to answer where it processes audio and how it shows that it is not recording, before it answers how many languages it speaks.

Cited cases

Who
iFLYTEK (科大讯飞), its own stall team at the 139th Canton Fair; foreign buyers named in the source: EDELIO (Argentina), a Brazilian buyer, Ramtin and his mother (Persian speakers) and Hasan (a Kazakh representative)
What they did
The stall team used iFLYTEK's AI glasses (讯飞AI眼镜, called GlassClaw on the AstronClaw architecture, 40 g) to serve foreign buyers on the floor: simultaneous face to face translation across 122 languages, accents and dialects, an array of 5 air conduction microphones and 1 bone conduction microphone with lip reading noise reduction to survive the hall, automatic capture of the talking points and a structured summary sent by e-mail afterwards. Second source, on the fair's first smart wearables zone, opened 15/04/2026: https://cnews.chinadaily.com.cn/a/202604/16/WS69e0a3faa310942cc49a7f7d.html
What came out of it
Buyers from Argentina, Brazil, Kazakhstan and other countries queued at the stall to try them: EDELIO said it would recommend them to friends, the Brazilian buyer asked to become a reseller and Hasan promised to present the device in Kazakhstan.
caifuhao.eastmoney.com
Who
Miaolente (米奥兰特) International Exhibition Group
What they did
Launched "AI慧展 1.0", a kit for exhibitors combining smart translation, buyer matching and, for the first time, dedicated AI glasses for real time interpretation across 128 languages, Vietnamese and Indonesian included
What came out of it
The company reports that the glasses cut communication friction on the show floor and let an exhibitor negotiate with a foreign buyer without a phone or an interpreter
www.meorient.com
Who
Independent reviewer on 什么值得买 (SMZDM), a Chinese consumer review platform
What they did
Tested real time translation on the Rokid Glasses (89 languages online, 6 offline) while walking streets, ordering in restaurants and shopping, talking at a fair and on a factory visit, and in business meetings, comparing it against phone apps and dedicated translator devices
What came out of it
Verdict: the glasses complement the phone rather than replace it, and are at their best for listening without taking your eyes off the conversation in short exchanges (claimed latency under 1 s), while conversations with several people, strong accents and noisy rooms still degrade recognition compared with a dedicated translator; the closing recommendation is to use the glasses to listen and the phone to show the translated text to the other person
post.smzdm.com
Who
C. Scott Brown, journalist at Android Authority
What they did
Hands-on test of Live Translation on the Ray-Ban Meta (English to French) after the global April 2025 rollout (English, Spanish, French, Italian)
What came out of it
Simple scripted sentences translated "quickly and accurately", but the system fell apart in natural conversation, his partner's laugh mid-sentence threw the translator into unintelligible output, and voice activation never worked reliably; he also noted that German is not supported, which limits his annual trip to Berlin, and concluded that the feature is no better than Google Translate on a phone
www.androidauthority.com
Who
Google, Shahram Izadi (VP of Android XR) and Nishtha Bhatia (Google product manager), on stage at Google I/O 2025
What they did
Live on-stage demo of an Android XR glasses prototype translating a conversation in real time, Izadi and Bhatia spoke Farsi and Hindi to each other while the glasses showed English captions, pitched internally as "subtitles for the real world"
What came out of it
The translation demo stopped working after roughly fifteen seconds, when Gemini stopped translating what the presenters were saying; Google still positioned live translation as a flagship feature of Android XR and Gemini
www.engadget.com
Who
Mentra (Mentra-Community), maker of the open source operating system MentraOS
What they did
MentraOS ships a "Live Translation" mini-app (open source, Apache-2.0) that runs on the Even Realities G1 and Mentra Live, detecting the speaker's language automatically and projecting the translated caption on the lens
What came out of it
Available in the MentraOS app store as one of the most used mini-apps alongside Live Captions; Mentra uses "see live captions, stream your view, talk to AI" as the system's core pitch
github.com
Who
Mentra-Community (open source contributors)
What they did
Independent open source app "Live Captions" for smart glasses running on MentraOS, offering real time captioning on the lens for deaf and hard of hearing people in more than 60 languages, with speaker differentiation and offline operation
What came out of it
Published and maintained as a public GitHub repository, usable on any MentraOS compatible glasses (Even Realities G1, Mentra Live, Vuzix, among others)
github.com
Who
XRAI, cofounded by Jacqui Press (Chief Brand Officer, profoundly deaf from birth)
What they did
The XRAI Glass software runs real time captions of speech on Nreal AR glasses (as well as phone, tablet and PC), identifying who is speaking, in 9 languages, aimed at deaf and hard of hearing people; the company announced translation between languages, tones of voice and accents as future features, and names hospitality, policing and education as target sectors
What came out of it
Launched in July 2022 as one of the first AR captioning products sold direct to consumers (from GBP 399.99 for the hardware, plus a subscription of up to GBP 49.99 a month)
patient-innovation.com
Who
Donna Clemens, assistive technology consultant at the Texas School for the Blind and Visually Impaired (TSBVI)
What they did
TSBVI assessed the on-device text reading and translation of the Envision Glasses for blind and low vision students; the Instant Text, Scan and Batch Scan modes read signs, menus and documents out loud, with support for several languages so that a sign or menu in a foreign language is read already translated
What came out of it
Clemens called the product "a game changing device, a glimpse into the future", reliable on standard printed text, but the voice commands remain English only and handwriting recognition can fail when the writing is hard to read
www.tsbvi.edu
Who
New York State Unified Court System (chief of operations Justin Barry)
What they did
A memorandum effective 20 July 2026 bans smart glasses with a camera, microphone or other recording technology, live translation and captioning glasses included, from every state courthouse, for litigants, lawyers, witnesses, staff and visitors
What came out of it
New York became the first US state to ban recording-capable AI glasses from all its state courts; anyone who wears prescription smart glasses has to bring a spare ordinary pair or hand the device to building security; the rule makes no exception for a deaf or hard of hearing visitor who depends on captions on the lens
natlawreview.com
Who
California State Legislature, Senate Bill (SB) 1130
What they did
The bill would make it a criminal misdemeanour to operate a wearable recording device, explicitly naming AI glasses with "high resolution cameras, always-on microphones and real time AI assistants", to capture audio or video of another person in a business setting with a reasonable expectation of privacy, without consent, and would also criminalise switching off the recording indicator light
What came out of it
As of August 2026 the bill was still pending, and would have to clear the full legislature and be signed by the governor; the proposed penalty is up to 1 year in county jail and/or a fine of up to USD 1,500 per violation
www.mondaq.com

Devices used

Talk to us

Back to the use cases