Skip to content

StructLDM

Learn 3D human generation from 2D images

Visit official site

What StructLDM is

The model is described as a structured latent diffusion model that learns 3D human generation from 2D images and enables controllable editing.

What you can do with it

  • Generate diverse view-consistent humans
  • Mix different parts through combination generation
  • Edit local clothing without mask conditions

Pros & cons

✓ What we like

  • Local clothing editing and 3D virtual try-on are listed
  • Pretrained models and sample data are offered

! What to watch out for

  • The workflow requires the SMPL model placed in a folder
  • Dependencies and assets must be downloaded first

FAQ

Where was the paper published?

The material names ECCV 2024.

Who proposed it?

Tao Hu, Fangzhou Hong and Ziwei Liu of S-Lab at NTU.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review