GaussianBody is a novel method for reconstructing dynamic clothed human models from monocular RGB videos using 3D Gaussian Splatting (3D-GS). The method addresses the challenges of non-rigid deformations and rich cloth details in dynamic human reconstruction by incorporating explicit pose-guided deformation. It uses forward linear blend skinning (LBS) to deform Gaussians from the canonical space to the observation space, and introduces a physically-based prior to regularize the transformation between the two spaces. The method also includes a split-with-scale mechanism to enhance point cloud density and a pose refinement strategy to improve texture quality. Experiments on the PeopleSnapshot and iPER datasets demonstrate that GaussianBody achieves state-of-the-art photorealistic novel-view rendering results with high-quality details and fast training speeds. Ablation studies validate the effectiveness of each component in the method.GaussianBody is a novel method for reconstructing dynamic clothed human models from monocular RGB videos using 3D Gaussian Splatting (3D-GS). The method addresses the challenges of non-rigid deformations and rich cloth details in dynamic human reconstruction by incorporating explicit pose-guided deformation. It uses forward linear blend skinning (LBS) to deform Gaussians from the canonical space to the observation space, and introduces a physically-based prior to regularize the transformation between the two spaces. The method also includes a split-with-scale mechanism to enhance point cloud density and a pose refinement strategy to improve texture quality. Experiments on the PeopleSnapshot and iPER datasets demonstrate that GaussianBody achieves state-of-the-art photorealistic novel-view rendering results with high-quality details and fast training speeds. Ablation studies validate the effectiveness of each component in the method.