Transforming Video into Interactive Worlds: Odyssey’s Revolutionary AI Model
London-based AI lab Odyssey has unveiled a research preview of their model that transforms video into interactive worlds, primarily targeting film and game production. This innovation hints at the potential emergence of a new medium in entertainment.
The interactive video generated by Odyssey’s AI responds in real-time to user inputs. Users can engage with it through a keyboard, phone, controller, or eventually voice commands, with the developers likening it to a primitive version of the Holodeck.
The AI has the capacity to render realistic video frames every 40 milliseconds. Thus, any interaction, like pressing a button or gesturing, triggers an almost instantaneous response, establishing an illusion of active influence over the digital environment.
According to Odyssey, the experience feels as if one is exploring a glitchy dream—unrefined and unstable, yet strikingly original. It’s essential to note that the visuals aren’t at the level of polished video games just yet.
Technical Innovations
What differentiates this interactive video technology from standard video games or CGI lies in what Odyssey calls a ‘world model’. Traditional video models generate entire clips simultaneously, whereas world models predict the subsequent frame based on the current state and user inputs, akin to how large language models anticipate the next word in a sentence but are considerably more complex due to the high-resolution video involved.
Odyssey describes these world models as action-conditioned dynamics models. With each interaction, the model examines the current state, the user’s action, and the preceding history to generate the next frame. This process results in a more organic and unpredictable experience, moving away from pre-determined actions.
Addressing Challenges
Creating this technology is fraught with challenges, notably maintaining stability over time. Generating each frame based on preceding ones can quickly lead to errors, known in AI research as "drift." To mitigate this, Odyssey has employed a "narrow distribution model." This approach involved pre-training their AI with general video footage and then fine-tuning it on a more limited set of environments. While this compromises some variety, it significantly enhances stability, preventing chaotic outputs.
The lab reports that they are making rapid progress on their next-generation model, which promises a broader range of pixels, dynamics, and actions. Operating this advanced AI in real-time entails considerable costs, estimated at £0.80-£1.60 (1-2) per user-hour, dependent on clusters of H100 GPUs across the US and EU. Although this cost might seem steep for streaming video, it remains economical compared to traditional film or game production expenses. Odyssey anticipates that these costs will further decline as models become more efficient.
Future of Interactive Video
New technologies have historically spurred unique storytelling forms, from cave paintings to current video games. Odyssey posits that AI-generated interactive video represents the next evolution of storytelling.
If successful, this could revolutionize various fields, including entertainment, education, and advertising. Envision training videos where skills can be practiced interactively or virtual tours that allow exploration from home.
While the current research preview is merely an initial stride towards this vision, it serves as an intriguing glimpse into the potential for AI-generated worlds to evolve into interactive experiences rather than remaining passive.
You can try the research preview here.
Discover the pinnacle of WordPress auto blogging technology with AutomationTools.AI. Harnessing the power of cutting-edge AI algorithms, AutomationTools.AI emerges as the foremost solution for effortlessly curating content from RSS feeds directly to your WordPress platform. Say goodbye to manual content curation and hello to seamless automation, as this innovative tool streamlines the process, saving you time and effort. Stay ahead of the curve in content management and elevate your WordPress website with AutomationTools.AI—the ultimate choice for efficient, dynamic, and hassle-free auto blogging. Learn More
