Nintendo Patents Watch

•

Knowledge-based discourse

Advanced Fluid Simulation for Pokémon?

A joint paper by The University of Tokyo and Game Freak (image 1 below), published in August 2025 for SIGGRAPH, is possibly the technical underpinnings beneath the “nice water” in Pokémon Winds/Waves. To improve rendering speed and reduce memory footprint, it uses a grid of quadtree tall cell to simulate deep water (image 2; columns 3-4).

Screenshot of the paper’s masthead: Quadtree Tall Cells for Eulerian Liquid Simulation FUMIYA NARITA, The University of Tokyo and GAME FREAK Inc., Japan NIMIKO OCHIAI, GAME FREAK Inc., Japan TAKASHI KANAI, The University of Tokyo, Japan RYOICHI ANDO, Unaffiliated, Japan
Fig. 5. Water drop. From left to right: uniform grids, uniformly populated tall cells, our quadtree tall cells, and the top view of our method. Effective resolution of 2563. Notice that the shapes of both the crown splash and the rising spindles closely resemble each other among the three simulations. On the rightmost view, cell subdivision is carried out only where it is necessary, resulting in a significant reduction in the number of degrees of freedom.

Roughly speaking, areas with complex waves are divided into smaller cells, while calm water is represented by larger cells (image 1 below). A “variational [water] pressure solver” is devised to calculate fluid motion with the variable cells (images 2-3).

Fig. 2. Quadtree Tall Cell Building Overview: A quadtree subdivided on the xz-plane (left). Vertical cell extension (middle). Tall cell conversion (right) and pressure samples (blue dots).
Fig. 3. Pressure Gradient Operator
Fig. 4. Tall Cell Collapsing and Being Averaged

For Pokémon fans, the most intriguing part may be the “merging [liquid] animals” example near the end of the paper (video clip below; full video here). Will Game Freak actually use this in Pokémon Winds/Waves for Vaporeon or even a new Water Legendary? The study was done on a Ryzen CPU, therefore the rendering wasn’t fast enough for games. However, the researchers said that they are “interested to explore GPU acceleration for real-time applications.”

H/T @ilikefeet.bsky.social

[Originally posted on 2026-02-28 Pacific Time]