A new generative model can build a 3D version of an entire city block or mountain range from a single prompt.
April 29, 2026
Original Paper
MetaEarth3D: Unlocking World-scale 3D Generation with Spatially Scalable Generative Modeling
arXiv · 2604.22828
The Takeaway
Generative AI has mastered 2D images, but 3D generation is usually limited to single objects or small rooms. MetaEarth3D is the first foundation model that can generate spatially consistent 3D environments at a planetary scale. It can create everything from detailed street-level views to massive, sprawling terrains that remain consistent as you move through them. This represents a massive leap in how AI understands and constructs physical space. This technology will likely change urban planning, video game development, and large-scale simulations.
From the abstract
Recent generative AI models have achieved remarkable breakthroughs in language and visual understanding. However, although these models can generate realistic visual content, their spatial scale remains confined to bounded environments, preventing them from capturing how geographic environments evolve across thousands of kilometers or from modeling the spatial structure of the large-scale physical world. This limitation poses a critical challenge for ultra-wide-area spatial intelligence in Earth