Join waitlist
Back to products
Odyssey

Odyssey

AI video you can both watch and interact with in real-time

Website odyssey.world
Overview

What it is

Odyssey is a new AI lab launching a research preview of real-time interactive video. Powered by a world model, it generates explorable video experiences you can influence.

Intent

I need it when

Develop intelligent robots that can operate in general, uncontrolled environments

Odyssey's world models enable robots to learn from video how humans manipulate the world, rehearse complex tasks like reaching and navigation before acting, and adapt to real-world variability—unlocking general-purpose robotic intelligence beyond brittle automation

Build next-generation interactive media and gaming experiences

Odyssey's world models power interactive games, immersive fitness software, art that responds to users, and interactive advertising by enabling real-time, open-ended simulations that users can control and influence within shared virtual worlds

Develop AI systems that understand and predict physical world dynamics

Odyssey's general-purpose world models are causal, multimodal systems trained on decades of video that learn to predict and interact with the world over long horizons, advancing the state-of-the-art in physical accuracy beyond language-only models

Create multi-agent collaborative simulations with human and AI participants

Agora-1 enables multiple participants—human or AI—to share and interact within the same world simulation in real-time, unlocking collaborative training, interactive experiences, and emergent multi-agent scenarios

Create interactive simulations for training and scenario planning

Odyssey-2 generates minutes-long interactive simulations from text or image prompts, enabling realistic edge-case scenario generation, warfighter training, employee upskilling, and personalized tutoring without waiting minutes for generation

Drop

Not a fit when

  • User needs traditional video generation or fixed-length video clips; Odyssey generates interactive, multi-minute simulations requiring real-time engagement
  • User requires immediate production-ready output; Odyssey is described as early-stage technology still in development
  • User needs deterministic, fully controlled outcomes; Odyssey's world models generate open-ended, interactive simulations with emergent behavior
  • User operates in highly constrained, non-visual domains; Odyssey is multimodal and learns from video, requiring visual or rich sensory input
  • User needs established, battle-tested commercial APIs; Odyssey is a research lab offering early API access to nascent models
Commercials

Pricing

Pricing not specified