TOM-GS Paper: Editable Video Representation via Pure Temporal Opacity Changes
kwangmoo_yi · x · 2026-07-29
The paper TOM-GS introduces a novel editable video representation. Unlike dynamic 3D Gaussian Splatting (3DGS) or Implicit Neural Representations (INRs) that rely on complex spatial deformations, this method forgoes physical movement, instead representing videos using pure temporal opacity modulation.
By assigning a learnable temporal mean and scale to the opacity of each static 3D Gaussian, the model allows spatial components to smoothly fade in and out of the scene. This approach maintains static spatial geometry, naturally supporting a wide range of manual and physics-based edits while ensuring seamless compatibility with established 3D editing tools.
Related event: TOM-GS Enables Editable Video via Temporal Opacity Modulation(2 posts)→
More from Multimodal
- Claude 5 Opus generates a textureless dirt-road car demo entirely on its own — ChrisGPT · 2026-07-29
- TILT improves compositional text-to-image generation with a model-intrinsic reward — Debottam Dutta · 2026-07-29
- Claude 5 Opus turns a no-texture dirt-road car demo into fully generated game graphics — ChrisGPT · 2026-07-29
- Stream3D turns frozen 3D generators into streaming models with bounded memory — pliang279 · 2026-07-29
- Creator turns Agent One into a 90-second cinematic horror trailer — LudovicCreator · 2026-07-29
- AI image experiment moved from realism to silkscreen after moiré issues — emollick · 2026-07-29