Odyssey's new AI model streams 3D interactive worlds

Odyssey, a startup founded by self-driving pioneers Oliver Cameron and Jeff Hawke, has unveiled a new AI model capable of streaming 3D interactive worlds directly from video.

This groundbreaking model generates and streams video frames every 40 milliseconds. This rapid generation allows users to explore areas within the video environment using basic controls, much like navigating a 3D video game.

The technology is based on a new "world model" designed to predict the next state of the world given current conditions and actions. Odyssey states this demonstrates key capabilities like generating realistic pixels, maintaining spatial consistency, learning actions from video, and producing coherent video streams for extended periods.

Currently available as an early demo on the web, the technology is still evolving. Odyssey acknowledges that the generated environments can appear blurry, distorted, and temporally unstable - meaning the scene might change unexpectedly as you move or look around.

However, the company is actively working on improvements, aiming for richer world representations that offer greater temporal stability and persistent state. They also plan to expand the model's capabilities beyond motion to allow for world interaction, learning from large-scale video data.

Odyssey employs a unique approach to data collection, utilizing a custom 360-degree backpack-mounted camera system to capture real-world landscapes. They believe this method provides a stronger foundation for higher-quality models compared to training solely on publicly available or synthetic data.

The startup has raised $27 million from investors including EQT Ventures, GV, and Air Street Capital, and counts Ed Catmull, a co-founder of Pixar, among its board members. Odyssey previously announced plans to enable creators to export and edit these AI-generated scenes in professional tools like Unreal Engine, Blender, and Adobe After Effects.

Odyssey envisions a future where "interactive video" replaces traditional video across various sectors, from entertainment and advertising to education and travel, fundamentally changing how we consume visual content.