AI startup Odyssey raises $9 million seed round for “Hollywood-grade” visual AI

Sora, which has so far been pitched to only a few early test customers but which OpenAI didn’t hide in a splashy demo in February, has Hollywood on edge. Tyler Perry, the film producer who came up with the Sora demo, said he was planning an $800 million film studio expansion in Atlanta “indefinitely” because Sora had made him doubt whether traditional studios had a future. Will there be a playground or not? ,

Actor and tech investor Ashton Kutcher has said that he also believes that AI will soon allow anyone to make a film without the need for actors, film crew or huge Hollywood budgets. Recently, another text-to-video style called Cling from Chinese AI company Kuaishou has also inspired AI developers.

Keep an eye on the beneficial grains

However so far, these models produce less than a modest scaled-down video sequence, and they cannot take into account that characters and backgrounds can be depicted consistently throughout the series. Odyssey seeks to give filmmakers more granular control over the images generated by AI style and lend itself to the stability that it does — as well as more affordable lighting and visual effects — and does so for longer frame sequences than the alternative. . According to Oliver Cameron, CEO of Odyssey and one of its two co-founders, AI models allow.

Cameron explained that Odyssey is designed as “Hollywood-grade” visual AI, aimed at empowering “the world’s most frenetic creators.” Luck, He cited the example of director James Cameron and his Avatar films, each of which took more than a decade to make, involved hundreds of visual effects artists, and cost millions of dollars. Odyssey wants to release that generation and cost without compromising the quality of the vision – potentially up to a team of 5 people spending $50,000 and working for 6 months.

Cameron says that to build right on the optical “Hollywood-grade” vision results of Odyssey, the equipment that Odyssey is developing is moving beyond a single text-to-video style. Rather, it’s fashion people focusing on four distinct tasks: creating impressive 3D renderings, creating content results, such as aqua or content, creating motion, and finally creating lighting fixtures results. They combine paintings to create frames of video. Moreover, the input to each of those models goes beyond textual content, he says. It will come with a human author drawing the drawings with a stylus or with virtual drawings.

This could allow filmmakers and professional visual effects artists – who are the intended customers of Odyssey’s tools – to have complete control over the overall output, something that is not possible with text-to-video flow AI fashion. .

Self-driving automotive giant

Cameron and co-founder Jeff Hawke, who is the noted generation executive of Odyssey, both came from the field of self-driving cars. Cameron co-founded the self-driving car company Voyage, which Cruise bought in 2021 for an undisclosed amount. He served as Cruise’s Product Vice President for the following 2 years. Hawke, meanwhile, was part of the initiation research and engineering team at British AI company Wave, which is building device brains for self-driving cars, which recently attracted a $1 billion investment from SoftBank, Nvidia and Microsoft. did.

After the leap from self-driving cars to tinsel, the city may no longer seem explorable, Cameron says At its heart, tools for self-driving cars will transform a 3-D world from two-dimensional input AI tool is ready. Vision results disease basically happens during the opposite – snatch a three-D global and condense it into two-dimensional shifted pictures.

At Vevey, Hawk had worked on an AI style called FIERY, which was the predecessor to another style called GAIA, that could snatch an initial video frame and the subsequent price, which could generate a complete, affordable street- Could develop quality video. This style immediately arose internationally. The text can be activated according to what is being shown in this simulated world. Wave is using that era to gather artificial intelligence to train its self-driving decision-making AI tools. However, Hawke was of the view that homogeneous concepts could be well adapted to create high-quality shot sequences for motion pictures.

Cameron says the optional benefit of coming out of the self-driving automotive sector is that both co-founders understand the utility of gathering their own real-world datasets to train their models.

Unlike alternative text-to-video companies that rely on scraping the web for video data to feed their AI tools, Odyssey is collecting its own personal 3-D datasets in the real world and using it to power its models. Is doing to feed. Cameron says there aren’t enough 3-D datasets available on the Internet to develop AI models of sufficient scale and capability for what Odyssey wants to collect. And his and Hawke’s experience in self-driving vehicles has given them some genius concepts about easy ways to store the real-world knowledge that Odyssey wants, even though Cameron declined to mention it on-the-record. Explained what some of these forms are.

Odyssey plans to expand the Flow group to 13 full-generation employees as well as capitalize on the money from its seed round to develop larger AI models, a valuable proposition that requires the company to scale up to vast clusters of graphics processing. Have right of access. The types of laptop chips most commonly used in gadgets, AI packages.

Correction, July 8: A previous version of this story incorrectly stated that Jeff Hawke worked on Wave’s GAIA AI style. He worked on a predecessor to GAIA called FIERY.

Subscribe to Fortune’s Upcoming to Government publication to receive weekly takeaways on easy ways to corner business. Connect remotely.