ClotheDreamer: Text-Guided Garment Generation with 3D Gaussians

TL;DR

ClotheDreamer introduces a novel 3D Gaussian-based method for generating high-quality, customizable garments from text prompts, enabling realistic virtual try-on and animation with improved quality and flexibility.

Contribution

The paper presents Disentangled Clothe Gaussian Splatting (DCGS), a new representation that allows separate optimization of garments and body, along with bidirectional SDS supervision and a pruning strategy for loose clothing.

Findings

Outperforms existing methods in garment quality and realism

Supports custom clothing templates and physical animation

Enables effective virtual try-on applications

Abstract

High-fidelity 3D garment synthesis from text is desirable yet challenging for digital avatar creation. Recent diffusion-based approaches via Score Distillation Sampling (SDS) have enabled new possibilities but either intricately couple with human body or struggle to reuse. We introduce ClotheDreamer, a 3D Gaussian-based method for generating wearable, production-ready 3D garment assets from text prompts. We propose a novel representation Disentangled Clothe Gaussian Splatting (DCGS) to enable separate optimization. DCGS represents clothed avatar as one Gaussian model but freezes body Gaussian splats. To enhance quality and completeness, we incorporate bidirectional SDS to supervise clothed avatar and garment RGBD renderings respectively with pose conditions and propose a new pruning strategy for loose clothing. Our approach can also support custom clothing templates as input. Benefiting from our design, the synthetic 3D garment can be easily applied to virtual try-on and support physically accurate animation. Extensive experiments showcase our method's superior and competitive performance. Our project page is at https://ggxxii.github.io/clothedreamer.