jwtcode/comfyui-diffbir

ComfyUI DiffBIR

Comfyui-DiffBIR is a comfyui implementation of offical DiffBIR.

---

TODO:

• 2024.05.10 ✅ repo create
• 2024.05.11 ✅ sampler implement
• 2024.05.11 ✅ stage2 load implement
• 2024.05.12 ✅ stage1 load implement
• 2024.05.12 ✅ stage1 tile implement
• 2024.05.12 ✅ simple load implement
• 2024.05.12 ✅ sampler update
• 2024.05.12 ✅ sampler advanced
• 2024.05.14 ✅ readme update
• 2024.05.14 ✅ basic sr workflow
• fr support
• bfr support
• dn support
• multi images support
• stage1 and stage2 sampler split

Visual Results

Blind Image Super-Resolution

low resolution image	super resolution image

Installing

Step 1: clone the repo and install the dependencies

To install ComfyUI-DiffBIR in addition to an existing installation of ComfyUI, you can follow the following steps:

1. cd custom_nodes
2. git clone https://github.com/ComfyUI/ComfyUI-DiffBIR
3. pip install -r requirements.txt
4. Restart ComfyUI

Step 2: download the pretrained model

put the model into ComfyUI/models/diffbir/

model	download link	Baidu Netdisk
bsrnet	BSRNet.pth
swinir_face	face_swinir_v1.ckpt
scunet_psnr	scunet_color_real_psnr.pth
swinir_general	general_swinir_v1.ckpt
v2	v2.pth
sd_v21	v2-1_512-ema-pruned.ckpt

Workflows

task	workflow
blind super-resolution	bsr_workflow.json
Blind Face Restoration
Aligned Face Restoration
Blind Image Denoising