AI Proving Ground DemoNVIDIA Blueprint: Digital HumansAI Assistants and AgentsDigital HumanAI Proving GroundAI & DataNVIDIA
Video
•
2
views
•
2:24
•
May 6, 2026
Building Nova, WWT's AI-powered digital human
WWT Senior Software Engineer Joshua Coffman walks through how Nova was built — from ASR and LLMs to RAG, TTS and real-time rendering — all running on a single workstation, offline capable.
WWT Senior Software Engineer Joshua Coffman provides a technical walkthrough of how Nova, WWT's AI-powered digital human, was designed and built.
This video breaks down the architecture, components and design decisions behind Nova, showing how WWT brings together multiple AI and rendering technologies to create a realistic, conversational digital human.
What this walkthrough covers
- The end-to-end pipeline powering Nova, from speech recognition (ASR) to LLM-driven responses and text-to-speech (TTS)
- How retrieval augmented generation (RAG) is used to ground responses and improve relevance
- The rendering layer that enables realistic facial animation and lip sync in real time, powered by Unreal Engine
- The frontend and streaming approach that enables interactive experiences