Animating flowing water in 3D Gaussian splatting scenes in real time
Make It Flow: Animating Water in 3D Gaussian Splatting Scenes
Computer Vision and Pattern RecognitionGraphics
Summary
Creating realistic moving water in 3D scenes is difficult because most 3D methods assume everything is still. The authors introduce a way to add smooth, flowing water motion effects to static 3D scenes without needing complex motion capture or physics simulation. They use image-based motion estimation to create a consistent 3D motion field that moves the water in the scene in a looping, natural way. This technique keeps scenes easy to view from any angle and runs quickly for real-time use.
What this means in practice
- •For game developers: Add visually realistic flowing water effects to static 3D game environments without costly dynamic capture or simulation.
- •For visual effects artists: Create seamless looping water animations in 3D scenes for films and virtual production with improved rendering efficiency and realism.
Authors
Kjell Keune, Florian Hahlbohm, Jannis Möller, Susana Castillo, Martin Eisemann, Marcus Magnor
Abstract
Explicit 3D scene representations such as 3D Gaussian Splatting (3DGS) have recently enabled high-quality, real-time novel view synthesis. However, these methods assume static scenes, limiting realism in the presence of dynamic phenomena ubiquitous in real-world environments. While coherent motion (e.g., humans or vehicles) can be captured and reconstructed, high-frequency effects such as flowing water remain intractable under casual capture conditions, where dense multi-view, high-speed acquisition is not feasible. We present a method to animate water in static 3DGS scenes, enabling visually plausible motion without requiring dynamic capture or physical simulation. Our approach estimates 2D optical flow from reference images or model renderings and lifts it into a consistent 3D Eulerian motion field. We propose a cinemagraph-inspired algorithm that applies this motion field to deform Gaussian primitives, producing temporally seamless looping animations. The resulting representation retains real-time rendering performance, significantly enhances perceptual realism, and preserves free-viewpoint navigation within the scene. We evaluate our method on a newly captured dataset of in-the-wild waterfall scenes and demonstrate clear improvements over prior work in dynamic Gaussian splatting and 3D cinemagraph generation in visual quality and computational efficiency.