Odyssey Launches AI-Powered Interactive 3D World Streaming
Odyssey, a startup founded by self-driving car pioneers Oliver Cameron and Jeff Hawke, has unveiled a groundbreaking AI model. This innovative technology generates and streams interactive 3D worlds in real-time, accessible directly through a web browser.
Currently available as an early demo, the AI model produces video frames every 40 milliseconds. Users can navigate these 3D environments with basic controls, offering an experience akin to exploring a video game.
How the AI Model Works
Odyssey's AI model predicts the next "state" of the world based on the current state, incoming actions, and a history of previous states and actions. This is detailed in their blog post. The model generates realistic pixels, maintains spatial consistency, learns actions from video, and outputs coherent video streams for extended periods.
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
— Odyssey (@odysseyml) May 28, 2025
The Future of Interactive Media
World models are a hot topic in the tech industry, with companies like DeepMind, World Labs, Microsoft, and Decart exploring their potential. These models are expected to revolutionize interactive media, including games and movies, and enable realistic simulations for various applications, such as robot training.
While the technology holds immense promise, concerns exist within the creative community. Some fear AI could displace jobs, as highlighted in a recent Wired investigation. Odyssey, however, emphasizes its commitment to collaborating with creative professionals, not replacing them.
Odyssey envisions a future where all video content becomes interactive, from entertainment and advertising to education and training.
Current Status and Future Development
Odyssey acknowledges that the current demo has limitations. Generated environments can appear blurry, distorted, and unstable. However, the company is committed to rapid improvements. The model currently streams video at up to 30 frames per second using Nvidia H100 GPUs, costing $1-$2 per user-hour.
The world played forward, by a model. On the one hand, it's calm and serene. On the other, it's chaotic and terrifying. I think the model nailed it in both cases. pic.twitter.com/vuszeL9p0V
— Oliver Cameron (@olivercameron) May 28, 2025
Odyssey is actively researching richer world representations for improved stability and persistence. They are also expanding the action space to include world interaction, learning from large-scale video data.
Unlike many AI labs, Odyssey uses a unique 360-degree, backpack-mounted camera system to capture real-world landscapes. They believe this approach will result in higher-quality models compared to those trained solely on publicly available data.
Odyssey has secured $27 million in funding and boasts Ed Catmull, co-founder of Pixar, on its board of directors. The company is also developing software to integrate its AI-generated scenes with tools like Unreal Engine, Blender, and Adobe After Effects, empowering creators with more editing capabilities.