Odyssey

Odyssey

Odyssey is an AI lab building general world models — systems that learn to simulate an environment rather than describe one. Odyssey's current model, Odyssey-2, turns a text or image prompt into a minutes-long interactive simulation that a person can move through in real time, with generation beginning in roughly fifty milliseconds. A second system, Agora-1, extends the idea to multiple agents in a shared world. Both are open to the public through demo endpoints, which is unusual in a field where most world-model work stays behind a paper.

Production credibility: Odyssey describes itself as pioneering causal, multimodal world models, and it publishes working demonstrations rather than only research: Odyssey-2 can be tried at experience.odyssey.ml and Agora-1 at agora.odyssey.ml. The company reports interaction beginning in about fifty milliseconds and simulations running for minutes rather than seconds. There is no published pricing, and the demos are the product surface today, so this is a lab to evaluate rather than a tool to deploy.

Key Features

• Odyssey-2 generating minutes-long interactive simulations
• Real-time interaction, with generation starting in around 50 milliseconds
• Text or image prompts as the entry point to a world
• Agora-1 extending world models to multiple agents
• Public demo endpoints for both systems
• Causal, multimodal approach rather than frame-by-frame video generation

Ideal Use Case

Teams exploring interactive media, game prototyping and simulation-based training who want to see where world models actually are today rather than read about them. It also suits researchers and product people benchmarking interactive generation against conventional video models.

How Odyssey differentiates

Runway, Luma and Pika generate a clip you watch; Odyssey generates a world you move through, which is a different technical problem and a different failure mode — coherence over minutes of interaction rather than seconds of playback. World Labs and Decart are the closest peers, and the practical difference is access: Odyssey puts both of its systems in front of the public. The honest caveat is commercial readiness. There is no pricing and no product tier, so today it is a demonstration of capability rather than something to build a business on.

FAQ

Q: What is Odyssey? A: Odyssey is an AI lab building general world models: causal, multimodal systems that simulate environments. Its Odyssey-2 model generates minutes-long interactive simulations from a text or image prompt.

Q: Can I try Odyssey? A: Yes. Odyssey-2 is available at experience.odyssey.ml and the multi-agent world model Agora-1 at agora.odyssey.ml.

Q: How fast is Odyssey-2? A: Odyssey reports that generation begins in roughly fifty milliseconds, so the simulation responds in real time as you interact with it.

Q: How much does Odyssey cost? A: There is no published pricing. The public demos are the current product surface.

Q: How is a world model different from a video model? A: A video model produces a clip to watch. A world model simulates an environment you act inside, so it has to stay coherent as you move through it rather than only for the length of a rendered shot.

tl;dr

Odyssey is an AI lab building general world models. Odyssey-2 generates minutes-long interactive simulations from text or image prompts with roughly 50ms response, and Agora-1 adds multiple agents. Both are publicly testable; there is no published pricing.