The model I think is fairly self explanatory.
Description
FAQ
Comments (2)
What's your preocess for making Manga style loras? do you just train them on full pages, or do you split them on individual panels?
I generally just use full images as they are, the only images I remove are non-relevant artwork such as credit pages or images that are backgrounds only or the few images that are of a small panel that's not a full pages worth of work. You usually don't have to remove much to clear up the dataset, but the stuff that doesn't need to be included is pretty intuitive I'd think.