Do these glasses show text or arrows in the lens?

No, and we would rather you heard it from us than discovered it after a board meeting. The eyewear platform this product is built on is a camera-and-audio frame: it sees, it hears, it speaks to the visitor through open-ear drivers, and it has no display in the lens. What that gives a site is an audio-first tour with visual recognition, which is what most interpretation plans actually ask for once they are costed. If your plan genuinely needs an optical see-through layer — labels, arrows, a reconstructed facade over a ruin — that is a different hardware family with a different weight, price, battery and supplier conversation, and we will tell you plainly that it is an ODM evaluation rather than sell you an audio frame and call it AR.

Will visitors walk off with them?

Some will, and any supplier who promises otherwise is selling you a fantasy. A guide glass is a higher-value item than an open-ear receiver, so the operation has to be designed for it: a deposit or a contact record at issue, a lanyard or retention strap as standard on hire fleets, a bay in your existing rental cabinet that charges and reports the unit as present or absent, and a session record that names the stop, language and duration. Sites normally pick one of three models — a refundable deposit, an identity or contact record where local law allows it, or an accepted shrinkage rate priced into the hire fee. Tell us which one and we configure the issue flow, the reporting view and the on-screen wording in each language to match.

Does the visitor need their own phone?

Not to walk the route. Pairing, language pack download and any account sign-in happen on a phone once, at the desk or on the visitor's own device before the tour starts; the recognition, positioning and narration then run on the glasses. What still needs a connection is free-form spoken questions against a cloud model, and content upload if your team does not use the browser console. Sites that cannot rely on visitors carrying a device — schools, care-home groups, high-security floors, loan wardrobes — run the same product from paired pre-configured units with a staff tablet as the setup point.

How does the tour know where the visitor is?

Four ways, usually mixed inside one site. Bluetooth beacons are the indoor default: a battery tag every few metres, a phone-free fix on the frame, and no change to your fabric. Visual matching uses the camera against the exhibits you have registered, which works where beacons do not want to go — behind glass, in a stone vault, outdoors on a heritage route — and needs good reference photography of each stop. GNSS geofencing covers open outdoor routes where satellite position is reliable enough to trigger a chapter. And a group chapter mode simply advances on a guide's button. We survey before quoting, because the honest answer depends on walls, sightlines and ceilings, and we would rather change the plan on paper than after installation.

Home
Products
AI Glasses
Vox Guide Tour Glasses

Vox Guide AI Tour Glasses — the Visit Runs on the Frame, Not in the Hand

A self-guided tour that follows where the visitor is looking. The camera recognises the exhibit, the frame works out which stop that is, and the narration arrives through open-ear drivers in the language that visitor chose — while their hands stay empty and their eyes stay on the object. The companion application on a phone covers pairing, language packs and content sync; it is not the tour. Built on the AI eyewear platform we already publish, for sites that want interpretation without a queue at a hire desk.

MOQ: 300 units on the shared eyewear platform · mixed colourways welcome · sample before any fleet decision

Updated 20 Oct 2026 · specifications per hardware revision

AI glasses with an everyday acetate frame photographed from the front — The guiding layer runs on our published eyewear hardware: an everyday frame a visitor will actually put on.

Reference photography of the eyewear platform. The guiding software, its console and the content screens are deployment-specific — we show them on a video call with your own route loaded rather than in a stock image.

At a glance

Recognition: Camera vs. your registered exhibits
Positioning: Beacon · visual · GNSS · chapter
Narration: Open-ear, per-visitor language
Phone role: Setup and sync, not the tour
Content: Your text, browser console
Records: Session, stop, dwell, language
Frame: Acetate, 3 colours, prescription-ready
Weather: IP54 splash and dust
Display in lens: No — stated up front
Certification: CE RED · FCC · RoHS 3
MOQ: 300 units
Sampling: 3–5 days stock, 7–15 configured

Send a Site Brief Download the interpretation brief Request a pilot sample

Tell us the route, the indoor or outdoor split and how hire is handled today. A person in the project team answers — usually the same day, always within one business day.

Why sites look at this at all

A handheld guide is a device the visitor has to operate

Every receiver fleet carries the same hidden costs: a desk to issue it, a form to sign, a number to press, a strap to adjust, a shell to wipe and a stock to reconcile. None of that is interpretation. The glasses remove the object from the transaction — a visitor puts them on, looks at something, and hears about it.

Attention stays on the exhibit. With a handset, the visitor is reading a screen and a label at the same time, which is why audio guide take-up and completion rates rarely match the effort put into the script. Looking at the thing while hearing about it is the closest a self-guided visit gets to a good human guide.

Recognition beats guessing where the visitor is. A beacon says “you are in this room”. A camera that matches the object in front of it says “you are asking about this bust, not the one behind it”, which is the difference between a tour that answers and a tour that lectures.

The language is chosen once, per visitor. Not per group, not per headset channel. A family of four can hear four languages from one stop, and your staff stop repeating the five-minute version in three tongues every hour.

What the site learns is different too. Stops reached, dwell before the visitor moved on, where the route was abandoned, language mix by hour. That is the data that changes a display, prices a premium route, or justifies a second entrance — written per session, not inferred from a paper sheet.

What we treat as standard, not as an upsell. The visible record indicator on the frame — no artwork or manual wording that hides it; a retention strap option on every hire unit; per-unit serialised session records your staff can reconcile; a browser content console your curators use without buying a licence per seat; offline-cached narration so a dropped network does not stop a tour; and a plain-language data note naming what is processed on the frame, what leaves it and who is the controller.

Ask any supplier of guide glasses these three questions: what happens to a tour when the venue network drops mid-stop, whose merchant or identity record covers a 300-euro item handed to a stranger, and what the console costs your team in year three. We answer all three in writing, and the answers arrive with the configuration rather than during commissioning.

Architecture

What runs on the glasses, what on the phone, what on your network

This is the table that decides whether the product works for your visitors, so it is stated plainly rather than left to a demo.

Function	On the frame	On the visitor's phone	On the venue side
Exhibit recognition	Capture and match against the image set registered for your stops	Not involved	The registered image set, curated per stop
Knowing which stop this is	Consumes the active positioning source and maps it to a stop	Optional assist where a phone fix is better than the beacon grid	Beacon layout, geofence drawing or visual anchors, set at survey
Narration playback	Plays the cached track for the stop, in the visitor's selected language	Language pack download at setup	Audio masters, transcripts and the stop map
Spoken questions	Capture, playback, and answers against content authored for the stop	Companion app as the account and settings surface	Cloud model calls for open-ended questions, with an authored-content fallback when there is no connection
Session records & analytics	Written per stop and synced when a link exists	Not required	Console view by site, route, language and hour
Issue, deposit, return	Serial identity per unit	Visitor record capture where your policy uses one	Your counter flow, your payment service, your cabinet bay reporting the unit present or absent

Said early

The AR question, answered before you have to ask it

There is no display in this lens. The frame sees, hears and speaks; it does not draw.

That is a deliberate statement rather than an omission, because the phrase “AR guide glasses” is doing a lot of work in this market and buyers are entitled to know which hardware family they are being quoted. Our published eyewear platform is a camera-and-open-ear-audio build: light enough to wear for two hours, cheap enough to run as a hire fleet, and robust enough that a visitor does not fear breaking it. For the overwhelming majority of interpretation briefs — trigger the right story at the right object, in the right language, without a handset — that is the correct tool.

If your plan genuinely needs pixels in the lens — reconstructed walls, wayfinding arrows, a caption track for deaf visitors, bilingual text over an object — that is an optical see-through build with a different weight, price band, battery expectation and supplier shortlist. We will say it is an ODM evaluation, tell you what we do not currently ship, and put you in touch with what the trade-offs look like. Selling you an audio frame and labelling it AR would cost you a bad launch and us a customer, in that order.

For accessible interpretation specifically: an audio-first frame plus your existing large-print and caption material covers a lot of briefs without an optical engine, and the transcript of every stop ships with the content pack as standard.

Specification

Full parameter table

Confirm against the datasheet that accompanies your quotation — the eyewear columns are published figures for the base platform, the guiding columns are set per deployment.

Eyewear base (as published for our AI glasses platform)
Frame	Acetate, three colourways, prescription-ready; your local optician fits the lenses and we supply the frame specification and lens-blank compatibility list
Camera	12 MP wide with photo and video capture, plus a visible record indicator on the frame that cannot be disabled by artwork or manual text
Audio	Open-ear drivers and a five-microphone array; the ear stays open to a guide, a child and a traffic lane
Connectivity	Bluetooth 5.3 with BLE, for pairing, beacon reception and the companion application
Durability	IP54 splash and dust — a drizzle and a cleaning cloth, not a boat deck or a hose
Battery	Published as about eight hours of mixed consumer use. A hire fleet does not rely on that figure: units charge between sessions in the bay of a G3 cabinet or on a dock, and we size the rotation against your opening hours instead of one long walk
Media and cache	Route content and narration are held on the unit so a dropped venue network does not stop a tour mid-stop
Display in lens	None. Optical see-through is a separate ODM evaluation
Certification	CE RED, FCC, RoHS 3, REACH SVHC screening, battery UN38.3 with SDS for air freight; UKCA on request. The revision-specific document list arrives with your quotation

Guiding system (configured per site)
Stops per route	Set by your interpretation plan; each stop carries triggers, narration masters, transcripts and an optional question set
Recognition set	Registered photographs per exhibit, refreshed when a display changes. Lighting and glazing are surveyed, because they decide the hit rate more than any processing figure
Positioning	BLE beacons indoors, visual matching against the object, GNSS geofencing on open outdoor routes, or scheduled group chapters on a guide's control — usually mixed inside one site
Languages	Not limited by the frame. What ships offline is a configuration decision; the supported list is confirmed per firmware revision rather than quoted from a poster
Question answering	Two tiers, and we keep them separate in writing: answers drawn from content your curators authored for that stop, and open-ended answers that need a cloud call and therefore a connection, a privacy notice and a reviewer's approval
Content console	Browser-based; your team edits text, audio, stop order and trigger geometry. We do not charge per curator seat
Reporting	Session, stop sequence, dwell, completion point, language, battery and fault per unit; export for your own analytics
Fleet operations	Serialised units, issue and return records, retention strap option, bay insert tooled for the glasses case from ten cabinets
Data protection	We state what is processed on the frame and what leaves it; the venue or operator is the controller for visitor data, we supply the DPIA input sheet, and EU-region hosting is available on request
Commercial	MOQ 300 units; samples 3 to 5 working days on stock builds, 7 to 15 configured, credited against production; production 25 to 35 days after sample approval; 12-month warranty with a 2% free replacement-parts allowance

Before you quote it internally

What we need from the site, and what we will not promise without it

Send these and a configuration plus a written price follows within one business day:

Floor plan or route mapIndoor and outdoor split, wall materials if known, ceiling heights in vaults and halls
Stop listHow many exhibits, how much of the narration already exists, in which languages
Visitor numbersPeak per hour, seasonality, group versus individual mix
Hire policyPaid or free, deposit or contact record, who is liable for a lost unit
Network realityWi-Fi coverage, guest VLAN rules, whether a wired drop reaches the galleries

What we will not claim, in writing:

Metre-level accuracyNot promised from a beacon grid alone; we measure at survey and tell you which stops need a visual trigger instead
Recognition through a vitrineReflection and angle decide this; we test with your photographs before you commit to a stop list
An end-to-end tour with no connection at allNarration and authored answers run offline; cloud question answering does not, and we say which is which
A loss-free hire fleetNobody can guarantee that; the deposit, contact and shrinkage models can
Certification beyond the stated scopeReports are per model and hardware revision, and the absent ones are named as absent

Send an interpretation brief

Get a configuration, a positioning plan and a written quotation

Describe the visit you want, not the device. You get a stop-and-route proposal, the positioning method per area, a fleet size derived from peak arrivals, and unit pricing by quantity.

Answered byA person in the project team, not an autoresponder
IncludesPositioning option per area, content authoring format, fleet and charging plan, revision-specific document list, spares price list, freight option
No account neededNo list purchase, no newsletter, no payment

Prefer to talk? WhatsApp the factory or write to wangkf@wingworld.cn.

Questions buyers ask about this model

Does it replace our audio guide receivers and cabinet?: No, and the best deployments use both. A receiver fleet is cheaper per unit, easier to issue to a hundred schoolchildren at once, and familiar to your staff. The glasses are the premium tier: a higher hire fee, a shorter queue because nobody is operating a device, and recognition rather than a button press. Most sites pilot a small glasses fleet beside the existing receivers and let the numbers decide the mix.
How many glasses does our site need?: Derived from peak arrivals and expected take-up, not from floor area — the same arithmetic we use for cabinets, with a bigger per-unit value, so the first deployment is deliberately small. Send arrivals per hour, the paid or free hire model and your languages, and we propose a fleet with the charging rotation that supports it.
Will it work in our rooms?: The honest answer comes from a survey. Vitrines, low light, tall vaulted ceilings and stone walls all change what a beacon grid and a camera can do, and we would rather adjust the plan on paper than discover it during commissioning. We start from your photographs and floor plan, then either confirm on a call or ask for a short site visit.
Who writes the narration?: Your curators, in whatever form it already exists. The console takes text, audio masters and a stop map; the same content package that serves a G1 player or a G2 receiver serves this product. Where a site wants generated first drafts, we treat them as drafts — a person approves every stop before it reaches a visitor, and we will not ship unverified claims about an object.
Can the visitor's phone be left at the counter?: For the tour, yes. Setup happens once — either with the visitor's own device or on a staff tablet for pre-configured units — and after that the glasses run the route. The exception is open-ended spoken questions that depend on a cloud call, which need a connection and a privacy notice your reviewer has accepted.
How is visitor data handled, and who is the controller?: You are the controller for your visitors' data. We supply a written statement of what is processed on the frame and what leaves it, the DPIA input sheet your legal review needs, and EU-region hosting on request for deployments in scope. Recognition runs against the exhibit set you registered, not against faces; if a build ever changes that, it will not be presented as the same product.
What does the console cost after year one?: Say the number out loud to whoever quotes you. Our content console is licensed with the deployment, with no per-curative-seat charge, and stop content stays editable by your team between projects. Export of session records is part of the same package, because data you cannot take out is a reason to quote us badly on year two.

Also in this line

Vox One AI Glasses

The consumer-facing build of the same platform, with its published specification and offline language pair.

View details

Front view of the self-service rental cabinet

G3 Rental Cabinet

Issue, charge and reconcile a glasses fleet from the bays you may already own — insert tooled from ten cabinets.

View details

Handheld AI guide player shown from the front

AI Audio Guides line

Cabinets, players, receivers and group sets — the interpretation stack this glasses tier sits on top of.

View the category

Want to see it on your own route?

Send five photographs of exhibits and a stop list — we will tell you which ones a frame can recognise, before you write a business case.

Send a Site Brief Browse products