I trained on a dataset of 300 images of not very high quality, mostly from JAV. The chilloutmix checkpoint was used for training. The optimal weight is 0.8, but it depends on the checkpoint. For some reason, it is much harder to get a decent result on some checkpoints, curiousmerge2.5D is the easiest one I tested, while chilloutmix, which was used for training, is more difficult. With high weight, have some problems with unexpected censorship and icons in the corners of images
Description
With high weight, have some problems with unexpected censorship and icons in the corners of images
FAQ
Comments (4)
很棒
You seriously train this onn pixelated images? Genious...
Yes, not the best idea, but I found a enough images only from JAV, which unfortunately is always censored(. With the right weight, there is a sufficient number of generated images without mosaics, so in the future, when I collect enough, I think I will try to improve LoRa on them.
@Beats can you use anime images? its easier to gather them as there are lots on danbooru etc. use tags: 1boy grabbing_another's_hair ponytail/twintails fellatio