--- title: "The Market Is Too Noisy, but the Glasses Still Know Who Is Speaking" date: 2026-11-04 category: AI in Action site: NeuroAI canonical: https://neuroai.site/a/na-ai-glasses-18-1104 language: en --- # The Market Is Too Noisy, but the Glasses Still Know Who Is Speaking > At a market abroad, the stallholder names a price and you do not understand, so you pull out your phone and open a translation app; by the time you have done that, the stallholder is already serving At a market abroad, the stallholder names a price and you do not understand, so you pull out your phone and open a translation app; by the time you have done that, the stallholder is already serving the next customer. The awkwardness of a language barrier often lies not in failing to translate, but in this chain of actions being too slow. If you could simply look up at the person and hear the translation in your ear, things would be far simpler. On November 2, the Light of the Internet Expo of the World Internet Conference opened in Wuzhen. According to Xinhua News Agency, more than a thousand new technologies were on show, and smart glasses drew long queues of visitors. Earlier, at the World Artificial Intelligence Conference in Shanghai, iFlytek's AI glasses were named the treasure of the show, moving translation from software onto the face. ## A Question About Price Stuck in Your Throat iFlytek did not get into translation on a whim. Translation devices, translation earphones and simultaneous-interpretation microphones have been a product line for years; the glasses simply take what the company has learned about listening and speaking and put it in a shell that sits closer to the body. According to the company, it packs lecture-style simultaneous interpretation, face-to-face translation, call translation and visual translation into a single frame. The difficulty of cross-language conversation often lies in the details. Two people facing each other have to take turns speaking and take turns looking down at a screen, so one complete sentence gets chopped to pieces; with several people talking at once, a phone is even less able to tell whom to listen to. Translation apps solve getting the words right, but not keeping the conversation smooth. The idea behind the glasses is to leave your hands free. You do not have to pull out a phone, and the other person does not have to wait: you look at someone, that person speaks, and the sound and the translation follow that direction. According to China Science Daily, staff at the scene described the ability as translating whoever you look at. That reads more like a product description than a laboratory measure, and need not be taken as something magical. iFlytek is also a translation partner for conferences of this kind. According to the company, it has provided multilingual simultaneous interpretation and transcription for many forums. The glasses on the stand and the invisible interpreting setup in the conference hall draw on the same accumulated skill in listening and translating. ## Six Ears and One Mouth To catch a sentence in a foreign language amid a crowd, algorithms alone are not enough. According to China Science Daily, iFlytek's AI glasses carry five air-conduction microphones, one bone-conduction microphone and lip-movement-based noise cancellation, with several signal paths working together. The air-conduction mics pick up ambient sound, the bone-conduction mic picks up vibrations along the head, and the camera watches how the lips move. Together, the three paths serve one very concrete purpose: locking onto the person who is speaking. Several people talking over one another, someone passing behind you, the clink of cups and plates are all pushed into the background. Lip-movement recognition adds information that pure hearing easily loses; even when part of the sound is covered, the mouth shape is still there, so the system has one more clue to follow. A visitor from abroad at the Wuzhen stand said the surroundings were actually quite noisy, yet the glasses still recognized the speech accurately, completed the translation and did so smoothly. That is one person's impression, but it points to the pain the product most wants to solve: not translating a sentence beautifully in a quiet room, but reliably catching a sentence in messy reality. ## Locking onto One Voice in the Noise An exhibition hall is the least courteous of test rooms. According to China Science Daily, amid the noisy crowd at the Wuzhen scene, the glasses could suppress about 90 dB of exhibition noise and pick up sound from eight meters away. It is worth noting that these figures come from a specific on-site demonstration; in a noisier or emptier venue, performance may not be quite the same. A body of about 40 grams is what makes it possible to wear them all day. However good the noise cancellation, if the glasses press on the bridge of the nose and make you want to take them off within half an hour, they will not enter real life. Fitting microphones, a battery and a display into the shape of an ordinary pair of glasses involves engineering trade-offs that actually test a maker more than adding one more feature does. ## A Hundred Tongues in One Pair of Glasses Language coverage is another threshold. According to iFlytek, the glasses support real-time translation across 122 languages and accents, and some common languages can be translated offline. The specification on JD.com's product page also lists 122 supported languages, so the two line up. Menus, road signs and instruction sheets can be translated with a glance. Offline use deserves a note of its own. In an underground passage with no signal, on a plane, or in a remote town, online translation is helpless; loading common languages onto the glasses locally is what makes them usable at any time. According to the company, Chinese and some foreign languages can already be translated offline, so not everything has to depend on the network. There is also a third-party assessment cited for the underlying technology. According to a report cited by the company, IDC's China AI Translation Technology Assessment 2025, iFlytek's AI translation technology ranked first in all eight evaluation dimensions; CNMO and other outlets, reporting on the new product, also mentioned six full scores in that report. The report is relayed by the company, so readers should simply keep its provenance in mind. ## From Translating Sentences to Getting Things Done What deserves more attention is that it does more than translate sentences. According to iFlytek, the built-in GlassClaw capability can automatically search and compare information based on what is in front of you, and, once confirmed, organize it into a document and send it out. Let your gaze rest on a foreign-language contract or an exhibition panel, and the system first reads the content, then gathers the key points. So translation has moved half a step from understanding a sentence to finishing a task. Meeting minutes, a teleprompter, and saving what you see as editable text are all packed into the same frame. It begins to resemble an assistant hanging on your face, rather than merely a faster electronic dictionary. Hearing clearly is only the first step. According to the company, behind these capabilities is the Spark speech foundation model, which helps the machine tell who is speaking and what is happening around it; the same technology has also gone into everyday products such as input methods, transcription tools and smart notebooks. The glasses are not an island, but one visible corner of a whole set of listening technology. ## An Organ Grown on Your Face Look back and the form translation takes has kept changing: from a thick dictionary, to an app on a phone, to a pair of glasses worn on the face. With each change, a person makes one fewer motion. One less time pulling out a phone, one less round of taking turns, and cross-language conversation comes closer to the ease of ordinary chat. Put a camera and microphones on your face and privacy becomes unavoidable. Which sounds stay local, which are uploaded, whether the view in front of you is stored: the maker has to say clearly, and the wearer has to have a choice. The closer the technology sits to the body, the more carefully this account of trust has to be settled. The boundaries need stating too. Translating whoever you look at is a description of a product capability, not a promise that holds in every setting; pickup range, battery life and offline languages all depend on the environment and the settings. What truly makes a technology stick is never the brightest line in the marketing, but a rushed morning when it simply does not let you down. Drive for a day abroad, wander through a market, talk through a small deal, and language will still be a nuisance. But when the translation sounds at your ear and you no longer need to look down for a screen, the nuisance shrinks a little. Translation is slowly turning from an app you have to open into a fairly unremarkable organ on your face. **Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.** --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-ai-glasses-18-1104 Free to quote with attribution and a link to the original.