TOM-GS: Editable Video Representation via Temporal Opacity Modulation of Static 3D Gaussians

2026-07-21

AI Summary

Existing editable video representations rely on complex spatial deformations or folded distributions, which hinder optimization efficiency and limit downstream editing flexibility. This work proposes a novel video representation based on static 3D Gaussians: by introducing learnable temporal mean and scale parameters to continuously modulate Gaussian opacities over time, it achieves smooth appearance and disappearance effects while maintaining fixed spatial geometry. By eschewing intricate deformations and confining appearance modulation solely to the temporal dimension, the method preserves geometric consistency and seamlessly integrates with standard 3D Gaussian Splatting pipelines and mainstream 3D editing tools. Experiments demonstrate that the proposed approach outperforms existing editable video representations in both visual fidelity and editing flexibility.

Abstract

While Implicit Neural Representations (INRs) and dynamic 3D Gaussian Splatting (3DGS) achieve impressive results in video processing, they often fall short of producing representations that are easily editable. Recent methods address this by introducing complex spatial deformations or folded distributions, which constrain optimization and reduce flexibility for downstream editing. In this paper, we introduce TOM-GS, an editable video representation that forgoes complex deformations in favor of regular 3D Gaussians equipped with a continuous temporal opacity formulation. By assigning a learnable temporal mean and scale to the opacity of each Gaussian, our model enables static 3D spatial components to fade smoothly in and out of the scene. Grounded by robust, off-the-shelf pose estimation, our approach maintains a static spatial geometry that naturally supports a wide range of manual and physics-based edits. TOM-GS outperforms prior editable video representations in visual fidelity, while its reliance on standard 3D Gaussians ensures seamless compatibility with established 3D editing tools.

Problem

• editable video representation
• 3D Gaussian Splatting
• temporal opacity modulation
• spatial deformations
• video editing

Innovation

• Temporal Opacity Modulation
• Editable Video Representation
• Static 3D Gaussians
• 3D Gaussian Splatting
• Continuous Temporal Modeling