A complete rework of an existing model that i've been trying to make work for a while. This time with all new images in the dataset, removal of male hands, complete restitches of the images from the hentai. I'm pretty happy with how this turned out and trying a new method to try and get around mosaics and blurry gentials. The new method seems to work although it still gets locked into certain positions for ??? scenes, it may mean I might need to also apply the same method to avoid it but that might come as a version 2 of the model.

Trigger word is "nogami_izumi" and also use tags like side_ponytail, red_eyes to get her to be more fleshed out.

Trained on 105 images of her and took a lot of hours to demosaic, edit, stitch, crop and photoshop so they were not blurry, since the hentai uses a lot of motion blurring in scenes.

Will try and make the other girls but they will have very limited datasets, so it won't turn out as good as Nogami.