CRM is a high-fidelity single-image to 3D textured mesh generative model. By integrating geometric prior into network design, it can generate six orthogonal view images from a single input image, and then utilize convolutional U-Net to create high-resolution triplanes. CRM further employs Flexicubes as a geometric representation, facilitating direct end-to-end optimization on the textured mesh. The entire model can generate high-fidelity textured meshes from an image within 10 seconds, without requiring optimization during testing.