Raw point clouds acquired by 3D scanning devices are often corrupted with noise and outliers. Thus, point cleaning can be seen as a crucial preprocessing step for high-quality mesh reconstruction. The removal of points is based on the 2D display in traditional approaches, which is less intuitive due to the limited spatial scalability. In this research, we developed a novel point cleaning approach that allows users to remove unwanted points in a virtual environment. One can mark noisy points as deleted with a transparent spherical tool attached to a VR controller, direct visual feedback is provided. Unlike desktop applications, higher rendering and processing capabilities are required in VR. Especially for the case of large densely sampled point clouds. Thanks to the advances in graphic devices, especially the access to the OpenGL Compute Shader, real time massive point cloud processing becomes possible. We manipulate VRAM buffers with carefully designed Compute Shaders in real time to support both rendering and labeling operations. We believe that our approach is versatile and easy to integrate. Our approach also offers significant potential for point cloud processing and annotations in VR.