Significant advancements are yet needed to realize mainstream realistic renderings of computer-generated scenes. In order to eliminate the laborious process of acquiring material reflectance through spectroscopic measurements or software-based procedural modeling, we propose a material authoring framework. This framework is designed to generate surface reflectance properties of a given object in the form of albedo, specular, surface normals, occlusion, roughness, and anisotropy maps, solving for BRDF models, all based on user-defined material keywords that describe an arbitrary object. The resulting material reflectance assets can be seamlessly integrated into a traditional PBR-based shader or neural rendering pipeline, capable of adapting to different object classes. To accomplish this objective, we introduce a comprehensive BRDF database that comprises 80 distinct objects belonging to 17 material classes. Each object within the database is captured under diverse lighting conditions and from various viewing angles. Our material authoring pipeline follows a two-step approach. Firstly, we introduce a UV-space generative model to synthesize desired textures of a given geometry based on user-defined material class labels. Subsequently, we develop a material inference network to decompose the synthesized textures into maps of reflectance properties to describe the BRDF model. The network is trained using the proposed BRDF database. We demonstrate the effectiveness of our proposed material authoring pipeline by rendering photo-realistic images with PBR-based industry shaders and neural shaders. Our framework facilitates high-quality material authoring and image-based inverse rendering. Additionally, we contribute to a BRDF database to further research in this field.