Revolutionizing Entertainment: Odyssey’s AI-Driven Interactive Video Model

In a groundbreaking advancement for the entertainment industry, London-based AI lab Odyssey has launched a research preview of a revolutionary model that transforms video into interactive worlds. This innovative technology not only focuses on film and game production but also hints at a completely new medium for storytelling.

The Future of Interactive Video

Odyssey’s AI model creates interactive video content that responds to user inputs in real-time, allowing users to engage via keyboard, phone, or controllers—and soon even voice commands. The team at Odyssey likens this experience to an “early version of the Holodeck,” suggesting a future where viewers can step into their favorite narratives.

Real-Time Interaction

One of the most impressive features of this technology is its ability to generate realistic video frames every 40 milliseconds. This rapid response time allows for an immersive experience where every button press or gesture influences the digital environment almost instantly, creating a sense of agency that traditional video cannot offer.

Exploring the Interactive Experience

“The experience today feels like exploring a glitchy dream—raw, unstable, but undeniably new,” Odyssey explains. While the visuals may not yet reach the polished standards of AAA games, the potential for innovation is immense.

Understanding World Models

What sets Odyssey’s interactive video apart from standard gaming or CGI technologies is its use of a “world model.” Unlike traditional video models that generate entire clips at once, Odyssey’s world model predicts the next frame based on the current state and user interactions. This is akin to how large language models predict the next word in a sentence, but far more complex due to the intricacies of high-resolution video.

The Mechanics Behind the Technology

At its core, a world model functions as an action-conditioned dynamics model. Each interaction relies on the current state, user action, and historical context to generate the subsequent video frame. This creates a more organic and unpredictable experience compared to standard gaming logic, which relies on pre-programmed responses.

Addressing AI Challenges

Developing this technology is no simple feat. One of the main challenges with AI-generated interactive video is maintaining stability over time. Odyssey has tackled this issue by employing a “narrow distribution model”—pre-training their AI on general video footage and fine-tuning it on specific environments. This approach sacrifices some variety for improved stability, ensuring a coherent user experience.

Future Developments

Odyssey is already making fast progress on their next-gen model, which promises a richer palette of pixels, dynamics, and user actions. The infrastructure required to run this sophisticated AI technology comes at a cost, estimated between £0.80-£1.60 ($1-2) per user-hour, utilizing clusters of H100 GPUs across the US and EU. While this may seem costly for streaming video, it’s considerably cheaper than traditional film or game production. Odyssey anticipates that these costs will decrease as technology continues to advance.

The Next Frontier in Storytelling

Historically, new technologies have birthed novel forms of storytelling, from cave paintings to books and video games. Odyssey envisions AI-generated interactive video as the next evolution in this ongoing narrative journey. This could revolutionize various fields, including entertainment, education, and advertising. Imagine interactive training videos that allow you to practice skills or virtual travel experiences that let you explore destinations from home.

A Glimpse into the Future

The current research preview serves as a proof of concept rather than a finished product, but it provides an intriguing insight into what the future may hold. As Odyssey continues to develop this technology, we might soon see interactive worlds that become playgrounds for creativity and exploration.

Curious to try the research preview? You can experience it here.

Join the Conversation

Learn more about AI and big data from industry leaders at the upcoming AI & Big Data Expo in Amsterdam, California, and London. This comprehensive event is co-located with other leading conferences, including the Intelligent Automation Conference and BlockX. Explore the future of technology and innovation!