This paper presents a memory-efficient approach for 3D Gaussian Splatting (3DGS), which significantly reduces the memory footprint of the method while maintaining high visual quality and rendering speed. The original 3DGS method requires a large memory footprint due to the high number of 3D Gaussian primitives, the use of spherical harmonics (SH) for directional radiance, and the precision required to store primitive attributes. The authors propose three main solutions to reduce memory usage: (1) a resolution-aware primitive pruning approach that reduces the number of primitives by 60%, (2) an adaptive adjustment method to choose the number of SH bands for each primitive, and (3) a codebook-based quantization method combined with half-float representation for further memory reduction. These methods result in a 27× reduction in overall size on standard datasets and a 1.7× increase in rendering speed. The method is demonstrated on standard datasets and shows significant reductions in download times when used on mobile devices. The authors also evaluate the method on various datasets and compare it with other state-of-the-art methods, showing that their approach achieves a better balance between memory usage, speed, and visual quality. The method is implemented in a WebGL framework, demonstrating 20–30 times faster download times on a mobile phone. The results show that the proposed method is effective in reducing memory usage without significant degradation in visual quality, making 3DGS more practical for applications requiring efficient memory usage and fast rendering.This paper presents a memory-efficient approach for 3D Gaussian Splatting (3DGS), which significantly reduces the memory footprint of the method while maintaining high visual quality and rendering speed. The original 3DGS method requires a large memory footprint due to the high number of 3D Gaussian primitives, the use of spherical harmonics (SH) for directional radiance, and the precision required to store primitive attributes. The authors propose three main solutions to reduce memory usage: (1) a resolution-aware primitive pruning approach that reduces the number of primitives by 60%, (2) an adaptive adjustment method to choose the number of SH bands for each primitive, and (3) a codebook-based quantization method combined with half-float representation for further memory reduction. These methods result in a 27× reduction in overall size on standard datasets and a 1.7× increase in rendering speed. The method is demonstrated on standard datasets and shows significant reductions in download times when used on mobile devices. The authors also evaluate the method on various datasets and compare it with other state-of-the-art methods, showing that their approach achieves a better balance between memory usage, speed, and visual quality. The method is implemented in a WebGL framework, demonstrating 20–30 times faster download times on a mobile phone. The results show that the proposed method is effective in reducing memory usage without significant degradation in visual quality, making 3DGS more practical for applications requiring efficient memory usage and fast rendering.