Our Research

Creating tomorrow’s technologies​, today.

[object Object]

Synthesizing Realism

Achieving photorealism in video sets a very high bar.


The creation of Artificial Humans involves several challenges, from photorealistic rendering to behavior modeling and generation. Capturing the dynamics and richness of human behavior is equally essential. Our research integrates digital advancements, computer graphics, and animation to build virtual humans that not only look real but also act and react in a compelling manner.


Realtime at Scale

A key component of interaction is speed. NEONs need to mimic response times that we naturally embody in conversations as humans. A few seconds delay and the whole experience can be ruined. The models that drive NEONs work at remarkable speed, but also operate at scales and costs that make NEONs viable for business.


Converse with Everyone

Realism is immaterial if conversational AI does not live up to the task.

Rich conversations

NEONs are built on top of our custom-designed conversational AI engine using LLMs to understand and produce rich, natural conversations while avoiding the pitfalls of reasoning with LLMs. We integrate advances across NLP and iterate new designs to allow NEONs to speak in a variety of languages, settings, and applications.


Reliable scale

Our efficient models and scalable framework allows NEONs to be deployed at scale. Our infrastructure is cloud native, with each application deployable on demand. Together with Samsung, we can achieve top performance whether with 1 NEON or 1,000.



Advance​ our collective understanding of the world we share


Realtime hyper-realistic generative AI-powered virtual beings

Realtime hyper-realistic generative AI-powered virtual beings




665 Clyde Ave

Mountain View, CA

United States


Copyrights © 2023 NEON. All rights reserved