Terrain Diffusion: A Diffusion-Based Successor to Perlin Noise
a day ago
- #Computer Vision
- #Procedural Generation
- #Diffusion Models
- Introduces Terrain Diffusion, a novel AI-based successor to Perlin noise for infinite, real-time terrain generation.
- Core innovation is InfiniteDiffusion, enabling seamless, boundless landscape synthesis with hierarchical diffusion models.
- Features include compact Laplacian encoding for stability across Earth-scale ranges and open-source infinite-tensor framework support.
- Demonstrates practical application of diffusion models in procedural world generation, offering coherence and controllability at planetary scales.