LiveX introduces AI City hosts with 3D navigation and multilingual help

Visitors looking for directions or product advice are the audience for LiveX’s latest AI offering: life-size digital hosts that converse in physical venues. The company introduced AI City at Smart City Expo World Congress in Barcelona on November 3, 2026, adding 3D navigation, product recommendations and multilingual translation to its host platform.

A conversation instead of a menu

AI City’s hosts are human-like avatars presented through kiosks. A visitor can approach one for navigation, product recommendations, event guidance or answers about the venue. SiliconANGLE describes a system that can see and hear the person in front of it, then respond conversationally with personalized assistance.

LiveX frames the offering against two alternatives: conventional kiosks with fixed menus, and general-purpose robots it says remain years from widespread deployment. Its proposed middle ground is conversational help delivered at a fixed venue. At the Barcelona conference, the company says attendees could interact without downloading an app, scanning a QR code or entering a password.

The expanded capabilities combine visual guidance with spoken assistance. Interactive 3D maps support location-aware navigation and discovery, while 3D product recommendations address shopping questions. Multilingual translation is another addition. LiveX’s announcement also introduces a Physical AI Host experience powered by Lyra, its software layer for coordinating requests.

An announcement built on existing hosts

The announcement builds on an existing installation base. LiveX says its hosts are deployed at more than 100 customer sites, naming PUMA stores, San José Mineta International Airport, SAP Center, El Camino Hospital, and Marriott and Hilton hotels. Those are company-reported deployments, rather than an independently measured result of the new capabilities.

The Physical AI Host also has a history before this announcement: LiveX says the first version debuted at NVIDIA GTC 2026. AI City brings that host concept together with the expanded mapping, recommendation and translation capabilities, rather than marking the first appearance of a LiveX host.

For venue operators, LiveX’s stated goal is to extend personalized assistance across locations while staff handle more complex requests. That is the company’s proposed division of work: hosts answer routine product, venue and event questions, with people available for problems requiring deeper attention.

NVIDIA computing, LiveX interaction models

Underneath the avatar, LiveX combines NVIDIA infrastructure with its own interaction models. Its company release, distributed by Business Wire and published on VentureBeat, says the platform runs on NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. It also uses the Metropolis VSS Blueprint to build visual AI agents, alongside Cosmos and a fine-tuned Nemotron model.

Other NVIDIA components have narrower roles. Retrieval and Reranker help find relevant information. Omniverse libraries can be used to build digital twins—virtual representations of places—for location-aware applications. LiveX describes its own models as trained for human-like interaction using experience and interaction data from live deployments.

• Omni Model handles understanding and responses across different types of input; Noise Management is designed to support interaction amid background noise.
• Visual Generation Model creates visual responses in real time, while Lyra coordinates capabilities to carry out user requests.
• LiveX Guardrail supplies safety and policy controls. The company says the combined models coordinate speech, expression and movement in real time.