StrandHead: Text to Strand-Disentangled 3D Head Avatars Using Hair Geometric Priors

Xiaokun Sun1, Zeyu Cai2, Ying Tai1, Jian Yang1, Zhenyu Zhang1*
1Nanjing University 2The Hong Kong University of Science and Technology (Guangzhou)
*Corresponding author
ArXiv Code Gallery
Teaser Image

We propose StrandHead, a text-driven framework for generating strand-disentangled 3D head avatars that feature high-fidelity facial details and strand-based hair. By accurately capturing the internal geometry of hair strands, our approach seamlessly supports flexible hairstyle transfer and editing, as well as physics-based rendering and simulation.

Results Overview

Strand-Disentangled Head Avatar Gallery

(Celebrity Examples)

(Ordinary People Examples)

Video results are loading slowly, please be patient.

Abstract

While haircut indicates distinct personality, existing avatar generation methods fail to model practical hair due to the general or entangled representation. We propose StrandHead, a novel text to 3D head avatar generation method capable of generating disentangled 3D hair with strand representation. Without using 3D data for supervision, we demonstrate that realistic hair strands can be generated from prompts by distilling 2D generative diffusion models. To this end, we propose a series of reliable priors on shape initialization, geometric primitives, and statistical haircut features, leading to a stable optimization and text-aligned performance. Extensive experiments show that StrandHead achieves the state-of-the-art reality and diversity of generated 3D head and hair. The generated 3D hair can also be easily implemented in the Unreal Engine for physical simulation and other applications. The code will be released for research purposes.

Method Highlights

pipeline

Compared to previous methods, StrandHead generate 3D heads with fine geometry and lifelike textures, as well as realistic textured strand-based hairstyles, and the entire framework does not require any hair training data.

Methodology

pipeline

Strandhead consists of two stages: (a) Under the constraints of the human-specific diffusion model and the FLAME-volving prior loss, StrandHead generates a detailed and reasonable bald head. (b) By introducing a differentiable prismatization algorithm, orientation consistency loss and curvature regularization loss inspired by hair geometric priors, StrandHead achieves diverse and realistic strand-accurate hair creation without any requiring hair training data.

Comparisons with Text-to-Head Methods

Comparisons with Text-to-Hair Methods

Haircut Transfer

Haircut Editing

Physics-Based Rendering and Simulation

BibTeX

@article{sun2024strandhead,
  title={StrandHead: Text to Strand-Disentangled 3D Head Avatars Using Hair Geometric Priors},
  author={Sun, Xiaokun and Cai, Zeyu and Tai, Ying and Yang, Jian and Zhang, Zhenyu},
  journal={arXiv preprint arXiv:2412.11586},
  year={2024}
}