Odyssey, a cutting-edge startup launched by self-driving visionaries Oliver Cameron and Jeff Hawke, is making waves with its innovative AI model that allows users to engage with streaming video in groundbreaking ways.
Currently available on the web as an “early demo,” this model impressively generates and streams video frames every 40 milliseconds. With simple controls at their fingertips, viewers can delve into video landscapes much like exploring a 3D-rendered video game world.
Odyssey explains that their model uses current scenarios, incoming actions, and historical data to predict future states of the world. This world model boasts capabilities like generating realistic pixels, maintaining spatial consistency, learning actions from video, and providing steady video streams for over five minutes.
The potential of interactive video is vast, attracting interest from various startups and tech giants such as DeepMind, Microsoft, and Decart. They foresee a future where world models transform media, creating dynamic games, movies, and training simulations. However, the tech comes with concerns. Reports have highlighted how companies like Activision Blizzard utilize AI, raising worries about job cuts in creative sectors. A study suggested that AI could disrupt over 100,000 jobs in U.S. film, TV, and animation.
Odyssey promises to collaborate with rather than replace creatives. The company envisions interactive video shifting traditional entertainment, education, advertising, and travel into dynamic experiences.
Though the current demo has its challenges – blurry and unstable environments – Odyssey is determined to improve rapidly. With the ability to stream at 30 frames per second using Nvidia H100 GPUs, the potential for advancement is high.
Looking into the future, Odyssey is focusing on richer world representations to enhance dynamics and stability. They’re working on expanding the action space to include comprehensive world interaction, aiming to learn from extensive video datasets.
Odyssey diverges from typical AI labs by using a unique approach: a 360-degree, backpack-mounted camera system to capture real-world landscapes, potentially leading to superior model quality.
The startup has secured $27 million in funding from investors like EQT Ventures and GV. Ed Catmull, co-founder of Pixar and former president of Walt Disney Animation Studios, lends his expertise as a board member.
In an exciting development, Odyssey is creating software that allows creators to import scenes generated by its models into tools like Unreal Engine and Blender for further editing. The future of interactive video seems closer than ever with Odyssey’s innovations lighting the way.
