Two self-driving car veterans are poised to make it big in Hollywood — not as actors, but as builders of the AI software upon which future blockbusters will be made.
They launched Odyssey, a San Francisco-based startup that competes with companies like OpenAI, Runway, Kuaishou, and Metaphysic to provide professional filmmakers and animators with the tools to create professional-quality films using generative AI.
Today, Odyssey has emerged from nearly a year of “stealth mode” operations with $9 million in seed funding from GV (Google Ventures, Alphabet’s venture arm), with participation from DCVC, AirStreet Capital, and more than a dozen individual early stage investors and angel investors.
The company's AI software is similar in some ways to OpenAI's Sora and Runway's Gen 3, which turn text descriptions of a scene or a sequence of shots into short, high-quality cinematic videos.
Sora has so far only been released to a small number of early testers, but its unveiling by OpenAI in a splashy demo in February has unsettled Hollywood: Film producer Tyler Perry said he was putting plans for an $800 million film studio expansion in Atlanta “on indefinite hold” after seeing the Sora demo, saying Sora made him question whether traditional studios would have any relevance in the future.
Actor and technology investor Ashton Kutcher also said he believes AI can enable anyone to make movies, without the need for actors, crews or big Hollywood budgets. Last month, another text-to-video model called Kling from Chinese AI company Kuaishou also impressed AI developers.
Fine-grained control
But so far, these models can only generate short video sequences of less than a minute, and can't guarantee that characters and backgrounds will be portrayed consistently throughout the entire sequence. According to Oliver Cameron, CEO and co-founder of Odyssey, Odyssey wants to give filmmakers more control and consistency over the shots its AI models generate, while providing more realistic lighting and visual effects, and do so over much longer frame sequences than other AI models can.
Odyssey is designed to be a “Hollywood-grade” visual AI, aimed at empowering “the world's most nerdy creators,” Cameron said. luckHe gives the example of director James Cameron and his Avatar films, both of which took more than a decade to make, involved hundreds of visual effects artists, and cost hundreds of millions of dollars. The Martian hopes to reduce that time and cost to perhaps a five-person team spending $50,000 to work on it for six months, without sacrificing visual quality.
To realize Odyssey's vision of “Hollywood-grade” visual effects, Cameron says the software it's creating isn't just a single text-to-video model. Instead, it's a family of models specialized in four different tasks: generating three-dimensional graphic renderings, generating material effects like water and cloth, generating motion, and finally generating lighting effects. These models work together to generate frames of video. Moreover, the input to each model is not limited to text, Cameron says. It could include something drawn by a human creator with a stylus, or a still digital image.
This allows the target users of Odyssey's software – filmmakers and professional visual effects artists – to have full creative control over the final output, something that's not possible with current text-to-video AI models.
Self-driving car veteran
Cameron and co-founder Jeff Hawk, Odyssey's chief technology officer, both come from the self-driving car world: Cameron co-founded self-driving car maker Voyage, which was acquired by Cruise for an undisclosed amount in 2021, and later served as Cruise's vice president of product for two years. Hawk, meanwhile, was part of the founding research and engineering team at Wave, a British AI company developing the software brains for self-driving cars that recently raised $1 billion in investment from SoftBank, Nvidia and Microsoft.
While the jump from self-driving cars to Hollywood might not be obvious, Cameron said that at its heart, self-driving car software is AI software that learns about the three-dimensional world from two-dimensional inputs. The problem with visual effects is essentially the inverse: condensing a three-dimensional world into a two-dimensional movie.
At Wayve, Hawke was working on an AI model called FIERY, a precursor to another model called GAIA, that could take the first video frames and use them to build an entire realistic street-level video of the world the model generates on the fly. You can also use text prompts to shape what's depicted in this simulated world. Wayve uses its technology to create synthetic data to train its autonomous driving decision-making AI software. But Hawke thought a similar idea could be used to create high-quality shot sequences for movies.
Cameron said another advantage of coming from the self-driving car space is that the two co-founders understand the importance of collecting their own real-world datasets to train models.
Unlike other text-to-video companies that have relied on scraping video data from the internet to feed into their AI software, Odyssey collects its own 3D datasets in the real world and feeds them into its models. Cameron says there aren't enough 3D datasets on the internet to build AI models that are big enough and powerful enough for what Odyssey wants to build. And Cameron and Hawk's experience with self-driving cars has given the pair some clever ideas about how Odyssey will gather the real-world data it needs, though Cameron won't talk publicly about some of the methods.
Odyssey plans to use the seed round to expand its current staff of 13 full-time employees and build larger AI models, an expensive proposition that requires the company to have access to large clusters of graphics processing units, the type of computer chip most often used in AI applications.
Correction, July 8: An earlier version of this article incorrectly stated that Jeff Hawke worked on Wayve's GAIA AI model. He worked on GAIA's predecessor, FIERY.
