[SIGGRAPH 2026 Posters] Tex-Shadow

Tex-Shadow: Synthesizing Textured 3D Shadow Art via 3D-Aware Diffusion Prior

Bumsoo Kim^1,†, Sanghyun Seo^1,*
1Chung-Ang University, Republic of Korea
^†Derived from his Master's thesis at CAU; now with Smilegate  ·  ^*Corresponding Author

📰 Note: our full code will be released until Sep. 2026.

---

Overview

Tex-Shadow is an optimization-based framework that leverages 3D-aware diffusion priors to automate the generation of high-quality textured 3D shadow art from sparse multi-view inputs — the first method to support both colorful anamorphic exhibitions and light-projected shadow art.

Pipeline

flowchart LR
    A["📷 N-view Images\nsemantic-free"] --> B

    subgraph S1["Stage 1 — 3D Reconstruction"]
        B["LoRA Customization\nZero123++ · MVDream"]
        C["Multi-SDS+ Loss\n+ Instant-NGP"]
        D["Marching Cubes\nmesh extraction"]
        B --> C --> D
    end

    subgraph S2["Stage 2 — Texture Refinement"]
        E["UV Unwrapping\nxatlas"]
        F["Inverse Texture Opt.\nnvdiffrast · Zero123"]
        E --> F
    end

    D --> E

    F --> G1["🎨 Anamorphic Art"]
    F --> G2["💡 Shadow Art"]
    F --> G3["🧱 Voxel · Brick · Game"]

Loading

Method

Tex-Shadow is a hybrid optimization framework with two main stages:

1. Customization for Semantic-free Inputs — A pre-trained 3D diffusion model is fine-tuned via LoRA on input image–pose pairs, enabling it to handle heterogeneous (semantic-free) inputs where each view depicts a categorically different subject.

2. Explicit 3D Reconstruction — The customized prior guides Instant-NGP via a hybrid Multi-SDS+ loss for 3D-consistent geometry. A subsequent UV-space inverse texture optimization step recovers fine-grained color details through a differentiable rendering pipeline.

The resulting mesh is exported to diverse representations: triangular mesh, voxel grid, density fields, and 3D point cloud.

Code Release Roadmap

• Stage 1 — LoRA fine-tuning for semantic-free 3D prior customization
• Stage 1 — Instant-NGP reconstruction with Multi-SDS+ loss
• Stage 2 — UV-space inverse texture refinement (code/refinement.py)

Acknowledgements

This project builds upon the following excellent open-source works:

• DreamGaussian — mesh extraction and texture refinement pipeline
• iFusion — Zero123-based diffusion guidance for sparse-view 3D reconstruction

Citation

@inproceedings{kim2026texshadow,
  title     = {Tex-Shadow: Synthesizing Textured 3D Shadow Art via 3D-Aware Diffusion Prior},
  author    = {Kim, Bumsoo and Seo, Sanghyun},
  booktitle = {SIGGRAPH '26 Posters: ACM SIGGRAPH 2026 Posters},
  year      = {2026},
  address   = {Los Angeles, CA, USA},
  publisher = {ACM}
}