From Openai’s 4o to stable diffusion, AI foundation models that create realistic images from text prompts are now abundant. In contrast, basic models that can generate a complete, coherent 3D online environment from a text prompt are simply emerging.
Still, the question is, when these models are readily available. Matthias Niessner, currently one of Europe’s most prominent AI 3D model researchers, took an entrepreneurial leave from the Visual Computing & AI Lab at the Institute of Technology Munich, and discovered that the startups are working in the region.
Niessner, formerly a co-founder of Synthesia and a realistic AI avatar startup valued at $2.1 billion, has garnered an unusually large seed round for the $13 million European startup. The round was led by prominent early stage investors in Europe (UIPATH supporters, Peak Games, etc.), with speed investment and participation from several well-known angels.
Its round size is even more impressive considering Spaitial doesn’t have much to show the world yet, other than the recently released teaser video showing how text prompts generate 3D rooms.
But then there is the technical team that Niessner assembled. Ricardo Martin-Brualla, who previously worked on Google’s 3D teleconfencing platform, is now called Beam. And David Novotney spent six years in Meta and led the 3D asset generation project from the company’s text.
Their collective expertise will give them a chance to fight in a space already containing some competitors with similar focus on photorealism. There’s Odyssey, which has raised $27 million and chased entertainment use cases. However, there is also World Labs, a startup founded by AI Pioneer Fei-Fei Li, which is already valued at over $1 billion.
Niessner thinks this is still mostly competitive compared to what exists in other types of basic models, but also regarding the “bigger vision” he and others are pursuing.
“I don’t just want to have a 3D world, I want this world to act like a real world. [let you] Do something in it and no one really cracks it yet,” he said.

Video games to the real world
No one is actually cracking, even if the demand for a photorealistic 3D environment is. The promise of “trillion dollars” opportunities from digital twins to augmented reality seems large enough to excite VCs, but it is ambiguous and multifaceted, and multifaceted enough to understand strategy for the market. The most obvious use case is video games creation, but these models include real-world use in fields such as entertainment applications, 3D visualization used for structures, and ultimately robotic training.
Niessner hopes to bypass the issue by granting developers a Foundation Model license and devising downstream applications for their specific uses. He also enlisted former Kazoo executive Lu Krogers, the fourth co-founder of Palo Alto while he was a visiting assistant professor at Stanford University.
One of the first tasks regarding Spaitial’s roadmap is to identify partners that can work with previous models compared to models that need to wait for higher quality.
“I want to work with at least some of my partners,” Niessner said.
Compared to other well-funded AI startups, Spaitial is making money on its agenda. But first, you need to spend it on both calculations and employment. In the latter case, its focus is not on quantity, but on quality. According to Niessner, “The team won’t grow into hundreds of people any time soon. It’s not just that it’s not happening, it’s not necessary.”
Instead, Niessner and his co-founders are working on creating larger, more interactive 3D spaces, for example, that glass can realistically crush. This unlocks what Niessner calls “Holy Grail.” This means you can enter the 10-year-old text and create your own video game in 10 minutes.
In his view, this ambitious goal is more achievable than what might actually look like a low-cost fruit, as most gaming platforms have strict control over what third parties can add. Users can create 3D objects. That’s as if Roblox could, of course, unless they decide to build it themselves. But by then, Spaitial may be busy replacing CADs instead. The next chapter in the 3D generation begins.
Source link