This paper aims to generate materials for 3D meshes from text descriptions.<br>Unlike existing methods that synthesize texture maps, we propose to generate<br>segment-wise procedural material graphs as the appearance representation, which<br>supports high-quality rendering and provides substantial flexibility in<br>editing. Instead of relying on extensive paired data, i.e., 3D meshes with<br>material graphs and corresponding text descriptions, to train a material graph<br>generative model, we propose to leverage the pre-trained 2D diffusion model as<br>a bridge to connect the text and material graphs. Specifically, our approach<br>decomposes a shape into a set of segments and designs a segment-controlled<br>diffusion model to synthesize 2D images that are aligned with mesh parts. Based<br>on generated images, we initialize parameters of material graphs and fine-tune<br>them through the differentiable rendering module to produce materials in<br>accordance with the textual description. Extensive experiments demonstrate the<br>superior performance of our framework in photorealism, resolution, and<br>editability over existing methods. Project page: https://zju3dv.github.io/MaPa