An experimental concept lora to create furry doujinshi pages. It works well for generating templates, actions, and conveying a scene surprisingly- however anatomy will probably get messed up in each panel.
This is unlikely to generate something coherent right now, but it's been fun to play with regardless; and syncs well with many artist style loras.
PONY: doujin, doujinshi, manga, black and white, monochrome, action lines
ILLUSTRIOUS: furry_doujinshi, comic panels, speech bubble, monochrome
The illustrious version will generate blank speech bubbles (most of the time) for you to input whatever text you want manually in photoshop or whatever you prefer. I manually edited text out of every image in the data set, except for FX and Kanji.
Careful training and image selection was used to avoid it generating censor bars without prompting for them; it can still happen but it's highly unlikely to.
I have been experimenting a lot with loras like this, as well as various other tools to get more functional comic panel generation in Pony.
Description
FAQ
Comments (6)
I wanted to post the Illustrious version but I don't have time to make many good example images right now.
Like all comic loras it's unstable af because AI just does not do comics well, but it works consistently enough that with some work you can make things that actually make sense.
Can it make images in color?
Yes, put monochrome and grayscale in negatives instead of positives and most if not all should be color.
Not trained with color as a focus though, but it still works fine.
@pihlawrkr738 cool
CivitAi was kinda enough to unpublish the Illustrious version, and not tell me, because the images "Were not generated". They were, but they had added text.
This would not be a problem except I was not notified this had happened, so idk how long this has been delisted. Will fix asap.
Fixed, left chat boxes blank this time.