Odyssey, a startup founded by self-driving technology pioneers Oliver Cameron and Jeff Hawke, has unveiled a new AI model capable of generating interactive, 3D-like video experiences in real time.
The system streams video frames every 40 milliseconds and lets users explore scenes with basic controls, much like moving around in a video game.
The model is powered by a newly developed 'world model' designed to predict what comes next in a virtual environment based on prior actions and context.
Introducing AI video you can watch and interact with, in real-time!
Powering this is a new world model that imagines and streams video frames every 40ms(!). No game engine in sight.
We call it interactive video, and it's free for anyone to try right now (GPUs permitting)! pic.twitter.com/QtADRXCQ8z
While the current demo, available online, still features blurred and unstable visuals, Odyssey claims it can maintain coherent video streams for over five minutes. Streaming performance reaches up to 30 frames per second using Nvidia H100 GPUs, costing $1-2 per user-hour.
Odyssey is positioning its platform as the future of interactive media, suggesting it could transform entertainment, education, advertising, and training.
Unlike some AI developers facing criticism for displacing creative workers, Odyssey pledges to work alongside artists and provide tools compatible with software like Unreal Engine, Blender, and After Effects.
To support development, the company built a 360-degree backpack-mounted camera to collect real-world footage, aiming to produce more realistic outputs than models trained solely on public datasets.
Backed by $27 million in funding and supported by Pixar co-founder Ed Catmull, Odyssey plans to significantly improve model stability and expand its action capabilities in the near future.
