AI is getting seriously creative, moving beyond just text and images to build entire virtual worlds you can actually explore. We’ve seen experiments like an AI-generated Quake demo and even AI-powered versions of Minecraft. Google DeepMind is also jumping in, building models to “simulate the world.” Now, a new startup called Odyssey, backed by Pixar co-founder Edwin Catmull, is launching its own take: something it calls “interactive video,” and they’re letting people try it out today.
Contents
Think of it as video you can literally walk around inside, dreamed up by AI in real-time as you move. It’s an early peek at what interacting with entirely generated environments could feel like, available now as a research preview.
What Exactly is ‘Interactive Video’?
Odyssey describes it as “video you can both watch and interact with, imagined entirely by AI in real-time.” The core idea is to let you move through a generated visual space from a first-person perspective. Unlike a traditional video game built with models and textures, this is attempting to generate the entire scene on the fly, aiming for something that looks more like the real world, or at least a blurry, dynamic version of it.
The company is pretty ambitious, hyping it as an “early version of the Holodeck.” However, they’re also upfront about its current state, acknowledging it feels like “exploring a glitchy dream — raw, unstable, but undeniably new.” And that’s part of the fun (and weirdness).
Stepping Inside: The Odyssey Preview Experience
Diving into Odyssey’s interactive video feels a bit like navigating a very smeary, constantly morphing version of Google Street View. You control your movement using standard keyboard keys (WASD) just like you would in many PC games. The preview offers a few different environments to explore, including a woody area with a structure, a shopping mall interior, and an outdoor space like a parking lot.
Each time you enter or switch environments, things look a little different because the AI is regenerating what’s in your view continuously. The visual quality is, for now, quite fuzzy and lacks fine detail. You get about two and a half minutes per session in the preview before it resets, but you can easily reload and jump back in.
Screenshot showing a blurry, AI-generated outdoor scene with trees and a distant cabin in Odyssey's interactive video preview.
Close-up screenshot of a blurry, abstract ground texture and foliage generated by Odyssey's interactive AI video.
The Tech Behind the Magic (and the Glitches)
How is this happening? Odyssey says it’s using powerful clusters of H100 GPUs – specialized computer chips designed for AI tasks – located in the US and Europe. The system takes your input (like moving forward) and the history of the frames you’ve seen, then the AI model tries to predict and generate the next frame you should see, streaming it back to you in real-time. The company claims this process can happen as fast as 40 milliseconds.
But real-time AI generation at this scale is still incredibly complex, and the current preview shows it. Objects often lack “collision” – you might walk through a wall or a house you expected to be solid. Environments can shift and change in unexpected ways; a doorway might suddenly turn into a solid brick wall as you approach it. Even standing still can be weird; the view sometimes slowly rotates or drifts on its own. It’s full of the kind of charming (or frustrating) glitches you see in early tech demos.
Pixar co-founder Edwin Catmull, who is on Odyssey’s board, didn’t give a specific timeline for when the image quality will improve. However, he told The Verge that Odyssey is at the forefront of this kind of AI work and that improvements are constantly being shared within the broader AI community. He acknowledges the images are “noisy” but sees this as something that can be solved by applying neural network filters later on.
Inside an AI-generated building environment in Odyssey's interactive video, showing blurry pillars and ceiling.
An AI-generated outdoor scene in Odyssey's interactive video, depicting a blurry street or parking lot view with buildings.
What Does This Mean? Potential and Pitfalls
Let’s be clear: Odyssey’s preview isn’t replacing your favorite video game anytime soon. The glitches and constantly changing world make it too unstable for traditional gameplay. It’s also not a replacement for watching a movie; the unpredictable morphing is too distracting when you just want to relax and watch a story unfold. It’s not really a good blend of the two mediums either, at least not yet.
However, even with its current roughness, messing around in the preview gives you a tangible sense of the potential. Seeing an environment being built around you as you move is a genuinely new experience. Given how rapidly AI technology is advancing, it’s not hard to imagine a version of this down the road that is much more stable, detailed, and responsive.
While Odyssey’s interactive video is definitely not the Star Trek Holodeck yet, and there’s a long journey ahead, this preview is a fascinating glimpse into a possible future where AI doesn’t just create static content, but builds dynamic, explorable realities in real-time.
