## \1‘s AI Breakthrough: A New Era in Interactive Video
In a significant leap for entertainment \1, London-based AI lab Odyssey has unveiled a research preview of its groundbreaking model that transforms standard video into interactive experiences. This innovation initially focuses on creating world models tailored for film and game production but has the potential to evolve into an entirely new medium of entertainment. With Odyssey’s AI model, users can interact with video content in real time using various inputs, including keyboard, phone, or voice commands. The team has likened their creation to an early version of the much-coveted Holodeck from science fiction lore.
## Real-Time Interaction: The Heart of the Experience
What sets Odyssey’s \1 apart is its ability to generate realistic-looking video frames every 40 milliseconds. This rapid response time allows users to influence the digital world almost instantaneously, creating an experience that feels immersive and dynamic. According to Odyssey, “The experience today feels like exploring a glitchy dream—raw, unstable, but undeniably new.” While the visuals might not yet match the polished quality of AAA video games, the potential for new storytelling formats is clear.
## The Technical Mechanics: Understanding World Models
To grasp the uniqueness of this AI-generated \1 technology, one must delve into the concept of a “world model.” Traditional video models produce entire clips in a linear fashion, whereas world models operate frame-by-frame, predicting subsequent frames based on the current state and user inputs. This approach is reminiscent of how large language models predict the next word in a sentence but is far more complex due to the high-resolution video frames involved.
\1 describes a world model as an “action-conditioned dynamics model,” which continuously adapts to user interactions. Each interaction feeds back into the model, allowing it to generate the next video frame based on learned patterns from countless videos. This organic and unpredictable flow is a departure from conventional gaming logic, where actions are typically pre-programmed.
## Overcoming Challenges: Stability in AI-Generated Video
Creating such a model is no small feat, particularly when it comes to maintaining stability over time. One of the primary challenges in AI-generated \1 is a phenomenon known as “drift,” where small errors in frame generation can quickly cascade into larger inconsistencies. To combat this, Odyssey employs a “narrow distribution model,” pre-training their AI on general video footage before fine-tuning it on a more focused set of environments. While this may limit variety, it enhances stability, ensuring that the generated content does not devolve into incoherence.
The company has reported fast progress on its next-generation model, which promises a richer array of pixels, dynamics, and actions.
## The Economics of Interactive Video
The cost of running this advanced AI \1 in real time is noteworthy, ranging from £0.80 to £1.60 (approximately $1 to $2) per user-hour, depending on the infrastructure, which relies on clusters of H100 GPUs across the US and EU. Although this may seem steep for video streaming, it is a fraction of the costs associated with traditional film or game production. Odyssey anticipates further reductions as the technology matures and models become more efficient.
## Envisioning the Future: A New Medium for Storytelling
Historically, technological advancements have spawned new forms of storytelling, from cave paintings to film and video games. Odyssey posits that AI-generated \1 could represent the next significant evolution in how stories are told. If this vision holds true, we may soon witness a transformation in entertainment, education, advertising, and beyond. Picture immersive training videos that allow users to practice skills in real time or virtual travel experiences enabling exploration from the comfort of home.
While the current research preview is just a stepping stone—more of a proof of concept than a polished product—it offers a tantalizing glimpse into the possibilities of AI-driven worlds as interactive playgrounds rather than passive viewing experiences. To explore this innovative research preview, you can try it out [here](#).
As we stand on the brink of what could be a revolution in interactive storytelling, the potential applications are limited only by our imagination. The future beckons, and with it, a world where storytelling is not just consumed but actively participated in.
