Workflow is in the Training Data section to your right, along with reference genital images to use for the reference field (I recommend image 00539.png). To crop and stitch together for manhwa, you will need this custom node: https://github.com/Norian11/comfyui_ManhwaCropping

Quick explanation of the two new nodes. With my workflow opened, on the most left you have a node to load the Reference Image, and to the right of it there is a node to load a manhwa panel. Once loaded you can drag click to make a square anywhere in the image (you can go out of the borders, if the censored area is in the center of the square occupying 50 percent of it is the best).

Then run the workflow. And to the right bottom you will see "Manhwa Stitch Save" which previews the image uncensored, if you dont like it just change seed and parameters until you like one, if you like it click "Save and Stitch" it will save the whole manhwa page with the uncensored part stitched. Then you can load that image and uncensor other section.

You can also check my Manga/Doujinshi Colorizer Lora: https://civarchive.com/models/2395415?modelVersionId=2693376

And my Doujinshi Uncensor Lora: https://civarchive.com/models/2395643

Trained on 1.8k triplets, uncensors Blank Censor from colored panels. This lora takes a censored square and a penis reference image (to maintain some consistency between patches), and outputs an uncensored version with the style of the reference.

It has problems:
Reference Image is required, and it has to be a penis in a similar orientation to that of the censored image. If it is in a too different orientation it will make inconsistencies. Cannot uncensor pussy correctly despite having plenty in dataset. To improve in next version.

Recommended Setting:

Strength: 1.5

Model: Distilled Klein 9B

Trigger: uncensorMangaKlein. (required to word, and with the point at the end)

Prompting: Just the trigger, but if you need more accuracy for bad results add tags caption.