Odyssey

Founded by Oliver Cameron in 2023

Generate real-time interactive world simulations from text or images

Cost

Demo

Rating

★ People love it

Time to value

Requires Expertise

You can use Odyssey to generate long-running, interactive video simulations from text or image prompts using AI world models. Their flagship model Odyssey-2 streams multi-minute simulations in near-real-time (50ms), unlike traditional video models that take minutes and output short clips. Agora-1 allows multiple humans or AI agents to share the same simulated world simultaneously. PROWL-1 uses reinforcement learning to improve world model performance through game environment exploration.

What Odyssey does

Prompt the model with an image or text to launch a live simulationStream a multi-minute interactive world simulation in real-timeHost multiple participants inside the same simulated world simultaneouslyUse RL agents to explore and stress-test simulation environmentsAccess world model capabilities via API for application developmentGenerate controllable, physically accurate video scenarios from text descriptionsTrain robotic systems by rehearsing manipulation tasks inside simulationsGenerates interactive video simulations in ~50ms, not minutesStreams multi-minute simulations rather than fixed 10-second clipsAllows multiple humans or AI agents to inhabit the same simulation via Agora-1Accepts both image and text prompts to start a simulationPROWL-1 uses RL agents to adversarially improve world model accuracyStarchild-1 learns from multimodal interaction beyond just visual observationAPI access available for developers to build on top of world modelsTrained on decades of video data to simulate physical world behavior