What is this?
This is a small collection of workflows for ComfyUI that is meant to serve a certain function. These "mini-flows", as I call them, are meant to be used separately from a whole, since each one serves a different function and have different outputs.
Technically, since this is ComfyUI, you could meld all of these workflows together if you want, but I wouldn't personally recommend it.
Currently, it includes the following workflows:
Watermark removal using GroundingDINO + Anzhc's YOLO model (works with complex watermarks)
Add Perpendicular merge re-creation in ComfyUI (for re-creating CyberFix-perp and equivalents)
How to use it?
Step 0: Download ComfyUI and ComfyUI-Manager
Step 1: Download your workflow of choice
Step 2: Extract the .zip file
Step 3: Import either the workflow image or the .json file
Step 4: Download all missing nodes from ComfyUI-Manager and restart
Step 5: Have fun!
Q&A
Q1. (Watermark Removal) In your txt2img workflow, you use CLIPSeg instead to remove watermarks. Why not use that here as well?
A: CLIPSeg works for really obvious watermarks, but also have a lot of false positives. In the txt2img workflow, I abuse the false positives to "reroll" certain parts of the image as well. For more accurate work, however, GroundingDINO + SAM is more accurate and can detect complex signatures and watermarks to remove them, and the false positive rate is lower.