User LoRA
Bìa bị gỡ: Vi phạm quy tắc nội dung
Tải lại ảnh bìa. Bạn vẫn có thể tìm kiếm và sử dụng mô hình này.
A Lora trained on images of doggystyle images with a behind (also kinda below) view. Since this is a more complicated pose/angle it can look weird, but the results are generally better than I thought they would be. Some images are using my Blood's Blend model and others are using a custom merge i have yet to upload, got lazy with the images so plenty are the same prompt and/or low res. weights 0.6 - 1 likes to gen anal. Can gen from behind and the from below version. should probably use dimensions around 512x512. trying do versions besides all fours usually doesn't turn out well
recommended tags: not really sure, doggystyle adjacent tags should help/work, other than that idk
from my testing it doesn't play well with some Loras/Lycoris, while others it's fine. adding too much to the prompt seems to mess it up. txt2image faces usually look weird
