Direct3D‑S2: A Framework for High-resolution 3D Rendering

1. Tutorial Introduction

Direct3D-S2 is a high-resolution 3D generation framework jointly launched by Nanjing University, DreamTech, Fudan University and Oxford University on May 26, 2025. The framework is based on sparse volume representation and innovative spatial sparse attention (SSA) mechanism, which greatly improves the computational efficiency of the diffusion transformer (DiT) and significantly reduces the training cost. Direct3D-S2 surpasses existing methods in both generation quality and efficiency, providing strong technical support for high-resolution 3D content creation. Related paper results are "Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention".

This tutorial uses a single RTX A6000 card as the resource. The 3D model generation time is about 5-10 minutes.

2. Project Examples

3. Operation steps

1. After starting the container, click the API address to enter the Web interface

2. Once you enter the webpage, you can use the model

If "Bad Gateway" is displayed, it means the model is initializing. Since the model is large, please wait about 1-2 minutes and refresh the page.

Parameter Description:

• SDF Resolution: Select the resolution of the generated 3D model.
• Simplify Mesh: Controls whether to simplify the mesh. If checked, the mesh will be simplified according to the Faces Reduction Ratio.
• Faces Reduction Ratio: Controls the face reduction ratio during mesh simplification.

How to use

4. Discussion

🖌️ If you see a high-quality project, please leave a message in the background to recommend it! In addition, we have also established a tutorial exchange group. Welcome friends to scan the QR code and remark [SD Tutorial] to join the group to discuss various technical issues and share application effects↓

Citation Information

The citation information for this project is as follows:

@article{wu2025direct3ds2gigascale3dgeneration,
  title={Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention}, 
  author={Shuang Wu and Youtian Lin and Feihu Zhang and Yifei Zeng and Yikang Yang and Yajie Bao and Jiachen Qian and Siyu Zhu and Philip Torr and Xun Cao and Yao Yao},
  journal={arXiv preprint arXiv:2505.17412},
  year={2025}
}