MeshMosaic: Scaling Artist Mesh Generation via Local-to-Global Assembly

TL;DR

MeshMosaic tackles the challenge of generating artist-quality meshes at high triangle counts by introducing a local-to-global autoregressive framework that assembles boundary-conditioned patches. Each patch is generated with boundary guidance and local-to-global features, using per-patch local quantization to $512^3$ to preserve detail while maintaining coherence across seams. The approach achieves meshes exceeding $10^5$ triangles with state-of-the-art geometric fidelity and strong user-preference performance on ShapeNet, Thingi10K, and Objaverse, illustrating its practical potential for game and film pipelines. Overall, MeshMosaic offers a scalable, patch-based paradigm for high-resolution 3D mesh synthesis that overcomes long-sequence bottlenecks and enables fine-grained detail recovery in artist-driven assets.

Abstract

Scaling artist-designed meshes to high triangle numbers remains challenging for autoregressive generative models. Existing transformer-based methods suffer from long-sequence bottlenecks and limited quantization resolution, primarily due to the large number of tokens required and constrained quantization granularity. These issues prevent faithful reproduction of fine geometric details and structured density patterns. We introduce MeshMosaic, a novel local-to-global framework for artist mesh generation that scales to over 100K triangles--substantially surpassing prior methods, which typically handle only around 8K faces. MeshMosaic first segments shapes into patches, generating each patch autoregressively and leveraging shared boundary conditions to promote coherence, symmetry, and seamless connectivity between neighboring regions. This strategy enhances scalability to high-resolution meshes by quantizing patches individually, resulting in more symmetrical and organized mesh density and structure. Extensive experiments across multiple public datasets demonstrate that MeshMosaic significantly outperforms state-of-the-art methods in both geometric fidelity and user preference, supporting superior detail representation and practical mesh generation for real-world applications.

Figures (21)

• Figure 1: MeshMosaic empowers scaling up artist mesh generation to more than 100k triangles by assembling boundary-conditioned local patches into cohesive, high-resolution meshes. It delivers flexible support over mesh density and ensures the faithful retention of intricate design details. Faces are assigned random blue colors to better illustrate the mesh layout.
• Figure 2: Comparison with existing state-of-the-art approaches, including both academic and commercial models. MeshMosaic achieved better quality with a smaller model size.
• Figure 3: Mosaic Art mosaic.
• Figure 4: DeepMesh Tokenizer.
• Figure 5: The pipeline of MeshMosaic. During inference, our method first applies PartField liu2025partfield to obtain semantic segmentation of the input shape. The input point cloud is then sampled according to the segmented patches and the original shape. Finally, our approach produces a clean, highly detailed mesh by assembling the generated patches.