AI-Powered 3D Portrait Pipeline

This research aims to validate and optimize traditional facial reconstruction workflows, establishing a technical pipeline for automated content generation.

High-Fidelity Facial Model Reconstruction

Evaluated multiple photo-to-3D modeling methodologies, establishing baseline models with Face Builder and refining anatomical details in ZBrush to engineer-grade precision.
Conducted specialized research on lens distortion compensation to improve modeling accuracy.

Retopology & Texture Optimization

Performed retopology adhering to MetaHuman specifications.
Developed a proprietary PBR texture extraction workflow from photographic sources, preserving pore-level skin details in Substance Painter.

UE5 Skin Shading

Developed SSS materials in UE5 and calibrated lighting/camera settings to achieve photorealistic facial rendering.

XGen Hair Grooming

Created XGen groom templates as AI training references, later implemented as UE5 hair assets.

The Final Model

Successfully completed a highly accurate MetaHuman facial model, establishing clear technical requirements and production guidelines for automated digital human generation.

Related work: Provide similarity evaluation for generative models