Odyssey, a startup founded by self-driving pioneers Oliver Cameron and Jeff Hawke, is developing an AI-powered tool that can transform text or an image into a 3D rendering.
The tool, dubbed Explorer, is similar in some ways to the so-called world models recently demoed by DeepMind, World Labs, and Israeli upstart Decart. Given a caption like “a Japanese garden, with rich, green foliage,” Explorer can generate an interactive, real-time scene.
Odyssey claims its tool is “particularly tuned” for creating photorealistic scenes. That’s largely a consequence of the startup’s technical approach; the AI powering Explorer was trained on real-world landscapes captured by the company’s custom-designed, 360-degree, backpack-mounted camera system.
Odyssey says that any scene generated by Explorer can be loaded into creative tools such as Unreal Engine, Blender, and Adobe After Effects and hand-edited. How? Explorer uses gaussian splats, a decades-old volume rendering technique capable of reconstructing realistic scenes. Gaussian splats are widely supported in computer graphics tools.
“While early, we’re excited to see the levels of 3D detail and fidelity Explorer can already achieve, and its potential for use in live-action film, hyper-realistic gaming, and new forms of entertainment,” Odyssey wrote in a blog post. “Although earlier in research, generative world motion, all in 3D, holds exciting promise to enable artists to generate and manipulate motion in new and more realistic ways, in addition to providing fine-tuned control that’s difficult to replicate in generative video models.”
Odyssey acknowledges that Explorer has several limitations today. The tool takes an average of 10 minutes to generate scenes, for example, and its scenes are relatively low in resolution — and not free of distracting visual artifacts.
But the company says that it has already seeded Explorer to production houses such as Garden Studios in the U.K. and a “growing group” of independent artists. Those interested in testing Explorer can apply on Odyssey’s blog.
Creatives may have mixed feelings about tools like Explorer — particularly those in the video game and film industries.
A recent Wired investigation found that game studios like Activision Blizzard, which has laid off scores of workers, are using AI to cut corners, ramp up productivity, and compensate for attrition. And a 2024 study commissioned by the Animation Guild, a union representing Hollywood animators and cartoonists, estimated that over 100,000 U.S.-based film, television, and animation jobs will be disrupted by AI by 2026.
But Odyssey says it’s committed to collaborating with creative professionals — not replacing them. To that end, the company on Wednesday announced that Ed Catmull, one of the co-founders of Pixar and former president of Disney Animation Studios, had joined its board of directors and invested in Odyssey.
“Generative world models are the newest and most unexplored major frontier in all of artificial intelligence,” Odyssey wrote. “We aspire to worlds that build themselves, that feel indistinguishable from reality, where new stories are born and remixed, where human and machine intelligence interact for fun or purpose. If all we ultimately achieve are incrementally better films or games, we will have fallen short.”
Cameron was previously the VP of product at Cruise, while Hawke was a founding researcher at Wayve. To date, Odyssey has raised $27 million from investors including EQT Ventures, GV, and Air Street Capital.