Munich-based startup Spaitial is developing a brand-new form of generative AI technology aimed at creating and understanding three-dimensional environments. This technology, known as Spatial Foundation Models (SFM), can generate both realistic and imaginary 3D scenes from text or images. Spaitial's goal is to enable AI with physical comprehension of space, a capability that is significantly lacking in existing AI models.
Spaitial’s SFMs focus on the generation and understanding of 3D structures, differing from traditional generative AI systems that only handle two-dimensional pixels or text. SFMs directly manipulate geometric shapes, materials, and physical properties, providing AI with deep comprehension of time and space. This physical understanding will lay the foundation for future world models, helping AI agents learn and navigate complex dynamic environments.
According to Spaitial, SFMs can generate photorealistic and physically accurate 3D scenes based on a single image, short video, or text prompt. Users can interactively explore these environments from any angle. By transforming outputs from flat 2D representations into spatially anchored 3D presentations, Spaitial aims to provide higher consistency, which is crucial for simulating physical processes or navigating dynamic real-world environments.
Spaitial sees broad application prospects, including immersive media such as gaming, film, augmented reality, and virtual reality, as well as digital twins for planning and simulation, automated robotics, and modeling industrial or urban infrastructure. For example, production facilities can be simulated in real-time, or autonomous machines can be trained to better understand their surroundings.
Recently, Spaitial successfully closed a $13 million seed round led by Earlybird Venture Capital, with participation from Speedinvest and several notable angel investors. These funds will be used to advance SFM technology, build infrastructure, and expand the team. The company also plans to collaborate with enterprises to test specific applications; interested individuals can join the early access waitlist.
The founding team of Spaitial has deep technical expertise and extensive entrepreneurial experience. Team members include Matthias Niessner, professor at Technical University of Munich and co-founder of avatar startup Synthesia, Ricardo Martin-Brualla, who previously worked on generating 3D technologies at Google, and David Novotny, an expert in digital twins from Meta. Additionally, Luke Rogers, who previously worked at McKinsey and Cazoo, brings commercial experience to the team. The team also includes Katja Schwarz, an early developer of GRAF and VoxGRAF, and Mojtaba Bemana, who won the 2025 EuroGraphics Doctoral Award for his paper "X-Fields."
Key Takeaways:
🌐 Spaitial is developing Spatial Foundation Models (SFM) to give AI physical comprehension of 3D environments.
🎮 SFMs can generate realistic 3D scenes from a single image or text, allowing users to freely interact and explore.
💰 The company secured $13 million in seed funding, planning to expand its team and technological applications while exploring market opportunities.