StdGEN++: A Comprehensive System for Semantic-Decomposed 3D Character Generation

TL;DR

StdGEN++ tackles the challenge of producing high-fidelity, semantically decomposed 3D characters suitable for production pipelines. It introduces a Dual-Branch Semantic-aware Large Reconstruction Model that separately handles global structure and fine facial details, coupled with a semantic-aware surface extraction formalism and a coarse-to-fine activation scheme to enable high-resolution meshes. A video-diffusion-based texture decomposition module provides editable, per-part textures (e.g., iris, skin) that support gaze tracking and downstream editing, while a unified input framework and an extended Anime3D-EX dataset underpin robust training. Extensive experiments show state-of-the-art geometric fidelity and semantic disentanglement, demonstrating practical benefits such as non-destructive editing, physics-consistent animation, and gaze control. Overall, StdGEN++ advances automated asset production by delivering structurally independent layers (body, hollow clothing, hair) and editable textures that integrate with industrial workflows.

Abstract

We present StdGEN++, a novel and comprehensive system for generating high-fidelity, semantically decomposed 3D characters from diverse inputs. Existing 3D generative methods often produce monolithic meshes that lack the structural flexibility required by industrial pipelines in gaming and animation. Addressing this gap, StdGEN++ is built upon a Dual-branch Semantic-aware Large Reconstruction Model (Dual-Branch S-LRM), which jointly reconstructs geometry, color, and per-component semantics in a feed-forward manner. To achieve production-level fidelity, we introduce a novel semantic surface extraction formalism compatible with hybrid implicit fields. This mechanism is accelerated by a coarse-to-fine proposal scheme, which significantly reduces memory footprint and enables high-resolution mesh generation. Furthermore, we propose a video-diffusion-based texture decomposition module that disentangles appearance into editable layers (e.g., separated iris and skin), resolving semantic confusion in facial regions. Experiments demonstrate that StdGEN++ achieves state-of-the-art performance, significantly outperforming existing methods in geometric accuracy and semantic disentanglement. Crucially, the resulting structural independence unlocks advanced downstream capabilities, including non-destructive editing, physics-compliant animation, and gaze tracking, making it a robust solution for automated character asset production.

Figures (12)

• Figure 1: Overview of the StdGEN++ pipeline. (1) Input Modality Unification: Diverse inputs (text or images) are first canonicalized into unified multi-view RGB and normal maps under A-pose. (2) Dual-branch S-LRM: These inputs feed into our reconstruction model, which leverages specialized Fullbody and Facial LoRA branches to predict high-fidelity geometry and semantic fields. (3) Surface Extraction: A semantic-aware extraction mechanism, accelerated by a coarse-to-fine proposal scheme, efficiently reconstructs high-resolution meshes from the implicit representations. (4) Texture Decomposition: Finally, the system performs texture decomposition to separate appearance components. Ultimately, the system yields structurally independent meshes (body, hollow clothing, hair) and editable texture layers.
• Figure 2: Demonstration of the structure and intermediate outputs of our dual-branch semantic-aware large reconstruction model (S-LRM).
• Figure 3: Our semantic-equivalent NeRF and SDF extraction scheme (shown in yellow color).
• Figure 4: Qualitative comparisons on geometry and appearance of generated 3D characters.
• Figure 5: Layered decomposition results. From left to right: input reference, generated holistic character, and semantically decomposed layers (body, cloth, hair). The cross-sectional views (rightmost column) reveal that the reconstructed clothing is accurately modeled with internal hollow structures, ready for physics simulation.