Recover interpretable dynamics from real splashing-liquid captures

Recover interpretable dynamics from real-world multi-view captures of splashing liquids by going beyond silhouette-based surface geometry and kinematic interface motion to model momentum-consistent fluid behavior, including transient structures such as impact bubbles and foam.

Background

SplashSplat reconstructs splashing liquids primarily from multi-view silhouettes. These observations constrain coarse surface geometry and interface motion, but do not reveal the specular glints and refraction that characterize water, and they miss transient structures such as impact bubbles and foam because such features are too small and short-lived to be reliably carved from the masks.

The method estimates a kinematic velocity field consistent with observed interface transport rather than enforcing a momentum balance. The paper therefore identifies recovery of physically interpretable dynamics from real captures as an unresolved direction that the proposed benchmark is intended to support.

References

Our velocity field is likewise kinematic: consistent with the observed interface, but not with a momentum balance. Recovering interpretable dynamics from real captures of this kind remains open, and is the direction this benchmark is built to support.

— SplashSplat: Reconstructing Splashing Liquids from Real-World Multi-View Videos  (2609.20818 - Liu et al., 17 Sep 2026) in Conclusion