Terrain Diffusion: A Diffusion-Based Successor to Perlin Noise in Infinite, Real-Time Terrain Generation
Alexander Goslin
2025-12-10
Summary
This paper introduces a new way to create realistic and incredibly large virtual worlds, moving beyond older techniques like Perlin noise which, while fast, don't look quite as real or maintain consistency over vast distances.
What's the problem?
Traditional methods for generating game worlds, like Perlin noise, are quick and can create endless landscapes, but they often lack the detailed realism we expect and struggle to maintain a consistent 'big picture' look when you zoom out to see a whole planet. They just don't create worlds that *feel* truly believable at a large scale.
What's the solution?
The researchers developed 'Terrain Diffusion,' which uses artificial intelligence, specifically a type of AI called diffusion models, to generate terrain. They created a new algorithm called 'InfiniteDiffusion' that allows this AI to create seamless, never-ending landscapes in real-time. They also use a system that combines a broad overview of the planet with fine details, and a clever way to store the information so it doesn't take up too much computer memory. Finally, they made the tools open-source so others can use and improve upon them.
Why it matters?
This is a big step forward because it means we can now realistically generate entire planets for games or simulations, with a level of detail and consistency that wasn't possible before. It opens the door to creating truly massive and immersive virtual worlds that feel more alive and believable, and it does so efficiently enough to be practical for real-time applications.
Abstract
For decades, procedural worlds have been built on procedural noise functions such as Perlin noise, which are fast and infinite, yet fundamentally limited in realism and large-scale coherence. We introduce Terrain Diffusion, an AI-era successor to Perlin noise that bridges the fidelity of diffusion models with the properties that made procedural noise indispensable: seamless infinite extent, seed-consistency, and constant-time random access. At its core is InfiniteDiffusion, a novel algorithm for infinite generation, enabling seamless, real-time synthesis of boundless landscapes. A hierarchical stack of diffusion models couples planetary context with local detail, while a compact Laplacian encoding stabilizes outputs across Earth-scale dynamic ranges. An open-source infinite-tensor framework supports constant-memory manipulation of unbounded tensors, and few-step consistency distillation enables efficient generation. Together, these components establish diffusion models as a practical foundation for procedural world generation, capable of synthesizing entire planets coherently, controllably, and without limits.