网上有很多整合包， 我个人喜欢纯净版， 自已搭建

1 拉代码

git clone https://github.com/comfyanonymous/ComfyUI 

如果没有装过git,下载安装:
https://git-scm.com/
https://git-lfs.com/

2 创建环境

cd ComfyUI
conda create -n ComfyUI python=3.11
conda activate ComfyUI

3 NVIDIA卡装pytorch, 各种包

nvcc --version 查看版本号, 要是运行不了来这里下载安装：https://developer.nvidia.com/cuda-downloads?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_local

关掉requirements.txt上的torch
查看显卡型号：cu121
安装torch: https://pytorch.org/get-started/locally/

pip install -r requirements.txt
conda list |grep torch    #查看是否cuda
pip install torch==2.1.2+cu121 torchaudio==2.1.2+cu121 torchvision==0.16.2+cu121 --extra-index-url https://download.pytorch.org/whl/cu121 xformers==0.0.23.post1 tensorboard==2.15.1 tensorflow==2.15.0
#--force-reinstall

4 启动

python main.py --windows-standalone-build    #--cpu

启动参数:https://www.yizz.cn/3943.html
启动时可能有各种问题，看控制台的错误提示

5 加节点，加模型

custom_nodes
models
运行节点时可能缺包，要监控控制台

参考：
https://blog.csdn.net/qq_28171389/article/details/143712138
https://blog.csdn.net/qq_28171389/article/details/143820731