A new paper on arXiv introduces Direct Conditional Parameterization for N-Dimensional Splatting, an extension of 3D Gaussian Splatting that incorporates additional conditioning variables such as view direction and time. According to the abstract, each primitive is lifted into a joint distribution over position and conditioning variables, allowing the representation to account for effects that vary with viewpoint or over time.
The work is motivated by the need to model view-dependent and dynamic effects, which are common in real-world scenes but not fully captured by standard static splatting approaches. By integrating these variables directly into the splatting formulation, the method aims to provide a more flexible framework for representing complex visual data.
The abstract is brief and does not include experimental results or comparisons with existing methods, so the practical performance of the approach remains to be evaluated. The paper is available on arXiv under identifier 2610.22778.