ComfyUI_ObjectClear

★ 27

图像编辑对象移除注意力机制ComfyUI插件

在ComfyUI中通过Object-Effect Attention实现完整对象移除，自动识别并从图像中消除指定物体，提高图像生成与修图的控制能力。

💡 在生成或导入图像中去除不需要的物体，便于精修与场景控制。

🍴 2 Forks💻 Python🔄 2025-11-24

🔗 GitHub 原文

📦

网盘下载

复制链接后前往夸克网盘下载

https://pan.quark.cn/s/c1eafc754fbb

📦 requirements.txt

accelerate
torch
torchvision
numpy
opencv-python
pillow
transformers
scipy
diffusers
segment-anything
matplotlib
gradio==5.30.0

📄 README

ComfyUI_ObjectClear

ObjectClear:Complete Object Removal via Object-Effect Attention,you can try it in ComfyUI

New

此方法优势在于内绘不仅限于主体，比如去除物体，物体的影子也会去掉，表现优异。

1 . Installation /安装

In the ./ComfyUI /custom_node/ directory, run the following:

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

2 . Requirements

通常不需要安装，因为没什么特别的库

pip install -r requirements.txt

3 . Model/模型

jixin0101/ObjectClear download files as below /下载下方列出的模型:

├── your comfyUI path/models/checkpoints/
|   ├──diffusion_pytorch_model.safetensors     #  unet dir/ 目录，rename or not  重命名或者随你，可以下fp16的
├── your comfyUI path/models/vae/
|   ├──diffusion_pytorch_model.safetensors     #  vae dir/目录，rename or not  重命名或者随你
├── models/clip_vision/
|   ├── diffusion_pytorch_model.safetensors    #  image_prompt_encoder 目录，其实就是这个： clip-vit-large-patch14.safetensors
├── models/clip/
|   ├── model.safetensors                                      # postfuse_module dir /目录 rename or not  重命名或者随你 ，可以下fp16 
|   ├── clip_l.safetensors                                        # normal comfyUI clip 常规comfy的clip
|   ├── clip_g.safetensors                                       # normal comfyUI clip 常规comfy的clip

seg /遮罩没有内置，随便用其他的吧。

4 . Example

5 .Tips/使用说明

支持单图或多图，遮罩合并用插件带的batch节点，512是垫图最小短边尺寸（会自动裁切），模型用512*512训练的，按理此尺寸是最好的。

Support single or multiple images, use batch nodes with plugins for mask merging, 512 is the input image ‘s minimum short edge size, and the model is trained with 512 * 512, which is theoretically the best size.

6. Citation

@InProceedings{zhao2025ObjectClear,
    title     = {{ObjectClear}: Complete Object Removal via Object-Effect Attention},
    author    = {Zhao, Jixin and Zhou, Shangchen and Wang, Zhouxia and Yang, Peiqing and Loy, Chen Change},
    booktitle = {arXiv preprint arXiv:2505.22636},
    year      = {2025}
    }