Narration that starts itself
Approach a point of interest and the story begins on its own — or speak only when it is needed. Quiet, Highlights, Full or Photo: the visitor decides.
Location triggered / four modesSolutionAI COMPANION / SYSTEM
FOR SCENIC AREAS, MUSEUMS
& CULTURAL DISTRICTS
Xiaoyou AI Companion brings interpretation, navigation, safety reminders and post-visit content together into one companion that is always on. The visitor simply puts on the glasses and speaks.
Talk to our team01The stack
Companionship is not one model's answer. It needs to know where the visitor is, what is in front of them, what the site wants to say, and what is happening on the ground. Four layers, none of them optional.
Voice is the first interface. There is nothing to learn — the visitor simply speaks. Audio stays at the ear, and the eyes stay on the landscape.
Continuous indoor and outdoor positioning, where geofences, the walkable network, POIs and safety reminders share one coordinate system. Location is not only for display — it informs narration and decisions.
Intent recognition, the knowledge base, route planning, image recognition and multiple languages work together around the visitor's position and question.
Content, devices, orders and data close the loop in a single console. Operators can see it, change it and keep learning from it.
02What it does
No capability stands alone. They gather around the same visitor, the same route and the same place, each aware of what the others are doing.
Approach a point of interest and the story begins on its own — or speak only when it is needed. Quiet, Highlights, Full or Photo: the visitor decides.
Location triggered / four modesContinuous indoor and outdoor positioning, with routes held to paths people can really walk. Directions use buildings and landmarks, not abstract bearings.
Indoor & outdoor / landmark cuesMulti-turn and continuous — no one has to repeat their question. It understands follow-ups, site-specific terms and accents.
Multi-turn / contextPhotograph what is in front of you and hear narration matched to that exact view. Recognition is not a trick — it keeps the answer where your eyes are.
First-person view / image recognitionSteps, water edges, restricted zones, closing hours. Nothing essential is missed, and nothing nags.
Geofencing / safetyRoutes, photos and narration clips are gathered into a travelogue or a short film, so the visit carries on after you leave.
Travelogue / imagery
03Open ecosystem
We do not manufacture glasses, and we do not rebuild foundation models. Current projects use Rokid Glass3; the hardware, operating system and foundation models come from ecosystem partners. What Xiaoyou builds in-house is the layer that makes a site actually work — the companion application, the spatial engine and the operations tools. Content is co-built with each site.
In-house scope: the companion application layer, the spatial engine, content tools and the operations console. Hardware, operating system and foundation models come from ecosystem partners.
Talk hardware and deployment04Delivery
We learn the routes, the space and the content assets, then agree the first phase and the real problems to solve.
We map the walkable network, POIs, geofences and positioning references, turning the site into a space that can be computed.
Scripts, the knowledge base and multilingual content are structured so they can be asked about, told well and translated.
Devices, console, voice and the on-site service flow are brought together and proven with real visitors, at a small scale first.
Content, routes and service keep evolving with real movement and feedback, so the system understands the site better the longer it runs.
05Let's talk
If you are looking for an AI companion service you can operate at a scenic area or a museum, talk to our team. We begin with your site and your questions.