Image: MTECH Digital - AI-generated original illustration — Original editorial illustration by MTECH Digital, created with AI assistance.
NewMedia’s AI HOST combined conversational AI, real-time animation and a large LED display for multilingual visitor interactions at IMEX America.
Key takeaways
- Four AI characters handled 402 conversations with 329 visitors.
- The installation supported English and 25 additional languages.
- Median response latency was reported at 2.6 seconds.
- Local fallbacks helped the experience continue during four connectivity outages.
NewMedia Creative Technology Studio’s AI HOST gave visitors at IMEX America 2026 spoken, real-time conversations with four animated digital characters. The project combined conversational AI with event-grade display, audio and real-time rendering rather than presenting the experience through a conventional chat interface.
How the experience worked
Visitors selected one of four personalities and spoke to it directly. NewMedia’s pre-event introduction described the characters as the NewMedia Guru, the Compliment Machine, the Oracle and the Quiz Master, each designed for a different style of audience interaction.
A post-event technical report from invidis said the installation used a 12-foot LED display, Unreal Engine 5 animation and AI infrastructure hosted in Spain. Speech recognition and response generation were processed remotely, while locally stored content and smaller models provided fallback behavior during connectivity interruptions.
Measured use on the show floor
NewMedia reported 402 conversations involving 329 visitors over four days, totaling more than 12 hours of dialogue. The system supported English plus 25 other languages, and the median interval between a visitor finishing a question and the character beginning its reply was 2.6 seconds.
Those figures were supplied by the developer and should be read as project-reported metrics rather than an independent performance audit. Even so, they offer useful operational detail that is often absent from experiential-AI demonstrations.
Why it matters
For museums, exhibitions and branded spaces, the lesson is that conversational AI is only one part of the experience. Microphone design, rendering, lip synchronization, display quality, network resilience and carefully defined character behavior all affect whether an interaction feels immediate and credible.
The project also shows a practical resilience pattern: pair cloud-based generation with local responses and models for degraded connectivity. Future deployments should additionally evaluate privacy notices, consent, accessibility, moderation and data-retention rules before collecting live audience speech.



