Odyssey, a startup founded by autonomous driving pioneers Oliver Cameron and Jeff Hawk, has developed an AI model that allows users to “interact” with streaming videos.
Available on the web in the “Initial Demo”, this model generates and streams video frames every 40 ms. Through basic controls, viewers can explore areas within the video, just like 3D rendered video games.
“When we consider the current state of the world, its next action, the history of nations and actions, the model tries to predict the next state of the world,” explains Odyssey in a blog post. “Powering this is a new world model, demonstrating the capabilities of generating pixels that feel realistic, maintaining spatial consistency, maintaining action learning from the video, and outputting coherent video streams for more than five minutes.”
Introducing real-time AI videos that you can watch and interact with!
Powered by this is a new world model that imagines and streams video frames every 40ms (!). The game engine is not visible.
We call it an interactive video, but anyone can try it for free right now (GPU is allowed)! pic.twitter.com/qtadrxcq8z
– Odyssey (@odysseyml) May 28, 2025
Many startups and large tech companies are chasing world models, including deep attitudes, influential AI researcher FEI-FEI Lee's World Lab, Microsoft, and Decals. They believe that using the world model for a day can create interactive media such as games and movies, and perform realistic simulations such as robot training environments.
However, creatives have mixed feelings about technology. Recent wired investigations have found that game studios like Activision Blizzard, which fired many workers, are using AI to cut corners and combat attrition. And a 2024 study commissioned by the Animation Guild, a union representing Hollywood animators and cartoonists, estimated that over 100,000 US-based film, television and animation jobs will be disrupted by AI in the coming months.
Odyssey is part of it and pledges to work with creative experts.
“Interactive video […] In his blog post, he wrote about the company, “I believe that entertainment, advertising, education, training, travel, and more will all evolve into interactive videos, Odes-driven interactive videos.
The Odyssey demo is a bit rough around the edge, and the company admits in its post. The environment the model generates is blurry and distorted, and unstable in the sense that the layout does not necessarily remain the same. If you move forward in one direction for a while or turn around, your surroundings may suddenly look different.
However, the company is committed to improving its models quickly. This model can stream video from a cluster of NVIDIA H100 GPUs at up to 30 frames, for a cost of $1-2 per “user hour.”
The world has advanced through models.
On the other hand, it is calm and calm. On the other hand, it's terrifying as chaotic.
In both cases, I think the model nailed it. pic.twitter.com/vuszel9p0v
– Oliver Cameron (@olivercameron) May 28, 2025
“Today, we are studying a richer world representation that captures dynamics much more faithfully, increasing temporal stability and sustained state,” Odyssey wrote in his post. “In parallel, we are expanding the action space from movement to global interactions, learning open action from large-scale videos.”
Odyssey takes a different approach than many AI labs in the world modeling field. We designed a camera system mounted on a 360-degree backpack to capture real-world landscapes. We believe this can serve as a basis for high quality models than models trained only with publicly available data.
So far, Odyssey has raised $27 million from investors such as EQT Ventures, GV and Air Street Capital. Ed Catmull, one of Pixar's co-founders and former president of Walt Disney Animation Studios, is on the startup's board of directors.
Last December, Odyssey said he was working on software that allows creators to load scenes generated by models into unrealistic engines, blenders, and Adobe After Effects, allowing manual editing.