NVIDIA Brings Physical AI to Life with LiveX AI

At GTC 2026, NVIDIA’s flagship conference for AI and accelerated computing, Jensen Huang came face-to-face with his own digital twin, Toy Jensen (TJ), powered by LiveX AI. Attendees could walk up and interact with TJ in real time, speaking naturally and hands-free in their preferred language.

NVIDIA featured TJ across 18 holographic installations throughout GTC. He could explain physical AI in French one moment and crack an AI joke in Italian the next, responding across 140 languages. He could help attendees find the room for a robotics session. TJ could also take a selfie for attendees, appearing right beside them in the photo.

Attendees interacted with TJ hands-free amid the noise of the conference. Understanding speech accurately and responding naturally in real time helped make those conversations feel intuitive.

Physical AI enables systems to interpret their surroundings and take action in the real world. LiveX brings this intelligence to people through Human-Like AI Interaction: an approachable presence that understands their needs and provides information and assistance. AI should adapt to people so that everyone can benefit from it, including those with limited technology expertise.

Approximately one in three GTC attendees interacted with TJ. Each attendee was invited to rate the experience immediately after their interaction. Those who provided feedback gave an average satisfaction score of 4.75 out of 5.

Attendees shared TJ selfies, group photos, and firsthand reactions across LinkedIn and other social platforms, sparking thousands of reactions and comments. Their posts carried NVIDIA’s physical AI story beyond the venue and into their own networks.

TJ’s journey continued at GTC Taipei and NVIDIA AI Day Singapore, bringing the LiveX-powered experience to more attendees across NVIDIA’s events. For Madison Huang, who tried TJ firsthand at GTC, the experience pointed to a broader possibility: making intuitive AI interactions part of everyday life.

In an AI City, LiveX’s Physical AI Hosts bring a human-like 3D presence to everyday places. These real-time AI agents understand people’s words, visual cues, and surroundings to offer recommendations, guidance, support, and moments of delight. People can speak naturally and hands-free with them, explore places and products in 3D, and create personalized experiences.

Powered by LiveX Interactive Intelligence and NVIDIA technology, these hosts combine what they hear, see, and know to understand people’s needs and take action to get things done. They can provide directions, take photos, create designs, or complete transactions such as purchases. Organizations can tailor the experience to their locations and services across different devices and physical settings.

For a shopper, AI becomes useful when it helps them discover something they might have missed. Dylan, PUMA’s AI Concierge, brings that possibility into the store. For travelers, LiveX has offered a welcome on arrival at SJC Airport, while guests at San Jose Marriott can discover local experiences and get guidance in their own language.

At NBA Summer League, fans put their imagination to work, designing hoodies and trying them on virtually. At El Camino Hospital, wayfinding helps people navigate an unfamiliar building. At San José State University (SJSU), LiveX brings an approachable AI presence into campus life, where people come to learn.

Illustrative applications of AI City powered by LiveX AI and NVIDIA.

For organizations, AI City expands the capacity to serve more people and meet more needs, from helping shoppers explore products beyond what is on display to supporting visitors across languages and locations.

LiveX Interactive Intelligence brings together proprietary models developed and trained for Human-Like AI Interaction, shaped by real-world interaction data and experience across live deployments. The LiveX Omni Model enables multimodal understanding and responses, while LiveX Noise Management helps people interact amid background noise. The LiveX Visual Generation Model creates visual responses in real time. Lyra (LiveX Your Real-world AI Agent) is the agent layer that coordinates these capabilities to carry out user requests in real time. LiveX Guardrail provides safety and policy controls. Together, the models coordinate speech, expression, and movement in real time.

Organizations can deploy LiveX in the cloud, in sovereign environments, or on premises, choosing the approach that fits their infrastructure and data requirements.

LiveX runs on NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, with lighter workloads optionally powered by NVIDIA GB10. The stack uses CUDA for GPU computing, TensorRT for inference optimization, and NVIDIA Triton for model serving. The platform uses a LiveX fine-tuned model based on NVIDIA Nemotron. NVIDIA NeMo supports retrieval and reranking to surface relevant information for each request. NVIDIA Cosmos supports environmental understanding, with world models that can serve as digital twins. NVIDIA Omniverse can also be used to provide detailed 3D digital twins for location-aware assistance.

Architecture: The Tech Stack Behind the AI City

AI City grows out of these everyday encounters. What begins as a useful answer, a warm welcome, or a moment worth sharing becomes something larger when it happens across a city: intelligence woven into daily life.

Powered by NVIDIA, LiveX is building the AI City block by block.

Humanoid robots can carry that intelligence too. The form may evolve; the focus remains on the person in front of it—understanding what they need and helping in a way that feels natural.

The AI City isn’t coming. It’s already open.

Intelligence. Live. In Person.