Mobile volumetric video with free-viewpoint navigation typically relies on 3D Gaussian Splatting (3DGS), where every Gaussian's attributes are encoded frame by frame using commodity video codecs. The authors of ResStreamGS argue that this all-Gaussians approach is inefficient for full-scene content, where large portions of the scene stay temporally stable. Their empirical study found that frame-wise adaptation and video coding of all Gaussians actually degrades quality in those stable regions.

To fix this, ResStreamGS uses a residual-sensitive dual-stream framework. It first identifies "residual-sensitive" Gaussians through lightweight probe adaptation. Stable Gaussians are transmitted once per Group of Pictures (GOP) and reused, while only the residual-sensitive ones remain in the frame-wise hardware-decodable video stream. This separates temporally invariant geometry from dynamic elements.

Experiments on complex full-scene volumetric videos show that ResStreamGS improves post-codec PSNR by 7.51–8.35 dB over a controlled V³-style all-Gaussians baseline. It also reduces amortized payload by up to 56.5% and achieves 30.883 FPS on a Qualcomm Snapdragon 778G platform, indicating the approach is practical for mobile streaming. The paper is a single source; no conflicting results are reported.