What if AI could generate an entire world while you're exploring it?
Not a video. Not a traditional game. An interactive environment that responds to your actions in real time.
Meet Odyssey-3, a new AI world model that generates environments from text prompts or images and lets you explore them as they evolve.
And yes, you can apparently become a seal, a cat, or someone kayaking through an AI-generated world. Walking like a normal human is also an option, although considerably less interesting.
What Odyssey-3 can do:
- Generate interactive 3D-like environments from text prompts
- Use images as starting points for new worlds
- Let you explore in first-person or third-person view
- Move the camera independently
- Introduce new events while the world is being generated
- Predict how objects move and interact using learned physics
But here's where things get interesting.
Odyssey-3 Pro scored 66.1 on Physics-IQ Verified, achieving the highest reported video-to-video benchmark result using best-of-8 sampling.
The benchmark tests how accurately AI models predict the continuation of real physical experiments, covering everything from fluid dynamics to optics and mechanics.
And Odyssey isn't building this technology just for entertainment.
The company is exploring applications in robotics, autonomous vehicles, and AI agents that need to understand and interact with physical environments.
I tried Odyssey-3 Flash in the browser.
The waiting time was around five minutes. Then I got roughly five minutes to explore before being sent back to the queue.
Five minutes to play God. Five minutes to wait. Apparently, even AI-generated universes have capacity limits.
Still, this is a fascinating glimpse of a future where worlds aren't designed in advance. They're generated as you experience them.
Try Odyssey-3:
https://experience.odyssey.systems/
Official announcement:
https://odyssey.systems/meet-odyssey-3
#Odyssey3 #AIWorldModels #GenerativeAI #InteractiveAI #ArtificialIntelligence
YouTube / ABVX