IMPORTANT: Use the V2 nodes (formerly Open3D). Open3D can crash on file save in the newest ComfyUI; use scipy in the V2 point cloud node.
Have not been able to pin point the issue due to differences in environments

No auto-downloads

This repo does not auto-download models or weights. If a model file is missing, the app will error out so you can provide it locally.

Install

pip install -r requirements.txt

TODO: Video node WIP (cant run at speed to test, has full debug logs for future help)
TODO: maybe using this for utils3d fixes the crashing (unsolved per-base issue):

git+https://github.com/EasternJournalist/utils3d.git@c5daf6f6c244d251f252102d09e9b7bcef791a38

Nodes

Use the V2 nodes (renamed from the Open3D variants).

Pixel Perfect Depth (Depth Map)
Pixel Perfect Depth (Depth Map - V2)
Pixel Perfect Depth (Video Depth)
Pixel Perfect Depth (Save Video Depth)
Pixel Perfect Depth (Unpack Depth)
Pixel Perfect Depth (Save Point Cloud)
Pixel Perfect Depth (Save Point Cloud - V2)

Models: where to place them

In comfyUI's models folder inside a subfolder named ComfyUI-pixel-perfect-depth Matching huggingfaces repo layout: https://huggingface.co/ApacheOne/ComfyUI-pixel-perfect-depth

Acknowledgement

We are grateful to the Depth Anything V2, MoGe and DiT teams for their code and model release. We would also like to sincerely thank the NeurIPS reviewers for their appreciation of this work (ratings: 5, 5, 5, 5).

Citation

If you find this project useful, please consider citing:

@article{xu2025pixel,
  title={Pixel-perfect depth with semantics-prompted diffusion transformers},
  author={Xu, Gangwei and Lin, Haotong and Luo, Hongcheng and Wang, Xianqi and Yao, Jingfeng and Zhu, Lianghui and Pu, Yuechuan and Chi, Cheng and Sun, Haiyang and Wang, Bing and others},
  journal={arXiv preprint arXiv:2510.07316},
  year={2025}
}