comfyui-old-photo-restoration

comfyui-old-photo-restoration
★ 61

旧照片修复图像修复深度学习ComfyUI扩展
在 ComfyUI 中本地执行旧照片修复（还原褪色、去污点、修补损坏），基于 Bringing-Old-Photos-Back-to-Life 论文，快速恢复历史影像细节。
💡 批量修复老照片、去除划痕与褪色并恢复细节。
🍴 9 Forks💻 Python🔄 2025-06-30
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/e00a65475347
📦 requirements.txt
PyYAML
dill
dlib-bin
dominate
easydict
einops
matplotlib
opencv-python
scikit-image
scipy
tensorboardX
📄 README

# ComfyUI Old Photo Restoration

This is an Extension for ComfyUI, which allows you to perform Bringing-Old-Photos-Back-to-Life natively.

example workflow

(You can drag this image directly into ComfyUI)

Original Paper: https://arxiv.org/abs/2004.09484

Original Repo: https://github.com/microsoft/Bringing-Old-Photos-Back-to-Life *(includes some example images)*

Installation

  • Install this Custom Node using the ComfyUI-Manager
  • Locate the installation folder of this Custom Node
    • cd ~ComfyUI\custom_nodes\comfyui-old-photo-restoration
  • Verify whether the lib_bopb2l folder exists or not
  • If not, download the backend.zip from Releases
  • Then extract the lib_bopb2l folder into this Custom Node

    • Prerequisite

  • Install as instructed above
  • Download global_checkpoints.zip from Releases
  • Extract and put the checkpoints folder *(not just the files)* into ~ComfyUI/custom_nodes/comfyui-old-photo-restoration/lib_bopb2l/Global
  • Download face_checkpoints.zip from Releases
  • Extract and put the checkpoints folder *(not just the files)* into ~ComfyUI/custom_nodes/comfyui-old-photo-restoration/lib_bopb2l/Face_Enhancement
  • Download shape_predictor_68_face_landmarks.zip from Releases
  • Extract the .dat file into ~ComfyUI/custom_nodes/comfyui-old-photo-restoration/lib_bopb2l/Face_Detection

    • The Releases page includes the original links, as well as the backups mirrored by myself

    Another mirror: Google Drive

    How to Use

    Simply connect an image to the Global Restoration node to process; if the image contains scratch artifacts, use the Global Restoration with Scratch Processing node instead.

    To improve the faces, connect the image to the Face Detection node then the Face Enhancement node, then connect the enhanced output and the image both to the Face Align node in the end to merge the results.

    Settings

  • HR: Use some higher parameters during the processings
  • GPU_ID: Specify the GPU to use
  • Set it to -1 to use CPU if you do not have a Nvidia GPU or are getting Out of Memory Error