Material Palette extracts a PBR material palette (albedo, normal, and roughness) from a single real-world image. It offers a method to map regions of an image to material concepts using a diffusion model, allowing for the sampling of texture images resembling materials found in similar scenes. Subsequently, an independent network decomposes the generated textures into spatially varying BRDF (SVBRDF), providing ready-to-use materials for rendering applications. This approach leverages a synthetic material library and a diffusion-generated RGB texture dataset to achieve generalization to novel samples via unsupervised domain adaptation. The product has been comprehensively evaluated using synthetic and real-world datasets, showcasing the applicability of its method for estimating materials from real photographs and utilizing them to edit 3D scenes.