OdysseyA boot found by Oliver Cameron and Jeff Hawke’s self-guiding pioneers has developed an AI model that allows users to “interact” with flow video.
Available online in an “early demo”, the model also creates video streams every 40mm of a second. Through basic checks, viewers can explore areas in a video similar to a 3D video game.
“Taking into account the current state of the world, an incoming action and a story of states and actions, the model tries to predict the next situation of the world,” explains the Odyssey in a blog post. “This supply is a new model of the world, which demonstrates opportunities such as creating pixels that feel realistic, maintaining spatial consistency, video learning actions and by extracting consistent video flows for 5 minutes or more.”
Some businesses and large technology companies are hunting after world models, including Deepmind, AI influence FEI-Fei Li’s World Labs, Microsoft and Decart. They believe that world models could be used one day to create interactive media, such as games and movies, and perform realistic simulations such as robot training environments.
But creatives have mixed feelings for technology. A recent Wired research has found that game studios such as Activision Blizzard, which has fired many workers, use AI to cut the corners and destruction of the battle. And 2024 study Assigned by The Animation Guild, a Union representing Hollywood animators and cartoonists, estimates that over 100,000 American films, television and cartoon jobs will be disturbed by AI in the coming months.
For its part, the Odyssey is committed to working with creative professionals – not to replace them.
“The interactive video … opens the door to completely new forms of entertainment, where stories can be created and explored at request, free from the restrictions and cost of traditional production,” the company writes in its blog post. “Over time, we believe that everything that is videos today – entertainment, ads, education, training, travel and much more – will evolve into interactive videos, all powered by the Odyssey.”
Odyssey’s demo is a bit rough around the edges that the company recognizes in place. The environments created by the model are blurry and deformed and unstable in the sense that their provisions do not always remain the same. Walk forward in one direction for a while or turn around, and the environment may seem suddenly different.
However, the company’s promise to quickly improve the model, which can carry videos to up to 30 frames per second from NVIDIA H100 GPU clusters with a cost of $ 1 to $ 2 per “user time”.
“Looking forward, we are investigating the richest global representations that record the dynamic much more faithfully, while increasing time stability and persistent situation,” the Odyssey writes in place. “At the same time, we are expanding the area of action from movement to global interaction. We learn open actions from large -scale videos.”
The Odyssey adopts a different approach from many AI laboratories to the world of modeling. He designed a 360 -degree camera system to capture real world landscapes, which he believes that the Odyssey can serve as a base for higher quality models than models trained exclusively in public data.
To date, the Odyssey has raised $ 27 million from investors, such as EQT Ventures, GV and Air Street Capital. Ed Catmull, one of Pixar’s co -founders and former president of Walt Disney Animation Studios, is on the boot board.
Last December, Odyssey said it is working on software that allows creators to load scenes produced by its models on tools such as Unreal Engine, Blender and Adobe After Effects so they can be manually processed.
