ComfyUI_DiffuEraser

DiffuEraser is a diffusion model for video Inpainting, running fully on the ComfyUI-native stack in this plugin.

Updates

• 改变封装方式，避免diffuser和transformer版本更迭带来的问题/ change the packaging method to avoid problems caused by the version updates of diffuser and transformer.

1. Installation

In the ./ComfyUI/custom_nodes directory:

git clone https://github.com/smthemex/ComfyUI_DiffuEraser.git

2. Requirements

pip install -r requirements.txt

3. Models

• VAE: sd-vae-ft-mse → models/vae/
• CLIP-l: any standard ComfyUI clip_l.safetensors → models/clip/
• PCM LoRA: PCM_Weights/sd15 e.g. pcm_sd15_smallcfg_2step_converted.safetensors → models/loras/
• ProPainter ProPainter address # below example
• BrushNet + motion UNet: lixiaowen/diffuEraser → models/DiffuEraser/

--  ComfyUI/models/vae
    |-- sd-vae-ft-mse.safetensors
--  ComfyUI/models/clip
    |-- clip_l.safetensors
--  ComfyUI/models/loras
    |-- pcm_sd15_smallcfg_2step_converted.safetensors
--  ComfyUI/models/DiffuEraser
     |-- brushnet
        |-- config.json
        |-- diffusion_pytorch_model.safetensors
     |-- unet_main
        |-- config.json
        |-- diffusion_pytorch_model.safetensors
     |-- propainter
        |-- ProPainter.pth
        |-- raft-things.pth
        |-- recurrent_flow_completion.pth

• Optional video-to-mask (RMBG / BiRefNet), auto-downloaded or local:

-- any/path/briaai/RMBG-2.0        # config.json, model.safetensors, birefnet.py, BiRefNet_config.py
-- any/path/ZhengPeng7/BiRefNet    # + handler.py

4 Example

• use single mask (old version)

5 Tips

• 对于水印，mask_dilation_iter（遮罩膨胀系数）应适当调低，比如2，常规使用Propainter的采样也就够了（最大边是960，应用了再超分吧）
• 现在12G也能跑1280*720，DiffuEraser的sample 节点的 blend支持2种输出，关闭为降低闪烁，开启为使用合成，避免loop循环的反复加载模型
• cfg>1.0 时 negative 生效

6 Citation

@misc{li2025diffueraserdiffusionmodelvideo,
   title={DiffuEraser: A Diffusion Model for Video Inpainting},
   author={Xiaowen Li and Haolan Xue and Peiran Ren and Liefeng Bo},
   year={2025},
   eprint={2501.10018},
   archivePrefix={arXiv},
   primaryClass={cs.CV},
   url={https://arxiv.org/abs/2501.10018},
}

@inproceedings{zhou2023propainter,
   title={{ProPainter}: Improving Propagation and Transformer for Video Inpainting},
   author={Zhou, Shangchen and Li, Chongyi and Chan, Kelvin C.K and Loy, Chen Change},
   booktitle={Proceedings of IEEE International Conference on Computer Vision (ICCV)},
   year={2023}
}

@misc{ju2024brushnet,
  title={BrushNet: A Plug-and-Play Image Inpainting Model with Decomposed Dual-Branch Diffusion},
  author={Xuan Ju and Xian Liu and Xintao Wang and Yuxuan Bian and Ying Shan and Qiang Xu},
  year={2024},
  eprint={2403.06976},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}

@article{BiRefNet,
  title={Bilateral Reference for High-Resolution Dichotomous Image Segmentation},
  author={Zheng, Peng and Gao, Dehong and Fan, Deng-Ping and Liu, Li and Laaksonen, Jorma and Ouyang, Wanli and Sebe, Nicu},
  journal={CAAI Artificial Intelligence Research},
  year={2024}
}