SAMT: Generating Structured Avatar Meshes and Textures from a Single Image

Despite rapid progress in generative 3D creation, producing high-fidelity 3D face assets from a single image remains challenging, as it requires both identity-critical facial micro-structures and fine-grained view-consistent textures. To address this, we present a two-stage framework named SAMT for monocular 3D avatar generation and texture synthesis. Specifically, a latent 3D diffusion model for facial mesh generation is pretrained and then further adapted to generate high-quality facial geometry through large-scale domain-specific finetuning on 35K curated 3D avatar models. Subsequently, the generated facial mesh is textured through a multi-view-aware texturing strategy. It incorporates multi-view facial priors along with the mesh geometry to guide a 2D texturing diffusion, enabling cross-view consistent and mesh-aligned texture synthesis. Extensive experiments demonstrate that SAMT improves over existing baselines by producing more coherent facial geometry together with more fine-grained and view-consistent textures. Project page is available at https://github.com/muyuWang/SAMT.

High-quality 3D digital faces are useful for virtual reality, games, and online communication, but creating them often requires special equipment or many images. A single portrait is much easier to provide, yet it only shows the face from one direction, so the unseen parts must be completed in a believable way. To address this challenge, we developed SAMT, a system that creates a complete 3D digital face from just one image. It first builds the overall 3D shape and then complete the unseen appearance, making the final face looks natural and consistent from different viewpoints. In our experiments, SAMT produced more natural face shapes and more coherent visual details than existing methods. This could make personalized 3D avatar creation more accessible for everyday applications.