A model for thick bottom waifus or semi-realistic characters. This model tends to produce hourglass figured ladies compared to my bottomheavy model.
Hyperfusion is the successor to this model, I suggest giving it a try
V3 Info:
v3 now uses LoRA! Edit: A1111 now supports LoRA models with the latest version. Use the following extension in Automatic1111 to load LoRA models https://github.com/kohya-ss/sd-webui-additional-networks
What changed since v2? It's the same dataset as v2, but trained with LoRA at a higher resolution of 768. Honestly this version was just a test to see how well LoRA training worked, but it turned out so well, I decided to release it.
Mixes incredibly well with AbyssOrange at 80-90% strength! Examples are all on AbyssOrange
Tips:
Place "huge" before "hyperass ass" to add extra emphasis
If you want front facing shots, dont use "hyperass ass", use "(thick thighs, wide hips)" instead
Description
v2 has a cleaner dataset, and better overall quality. Trained for much longer at a lower learning rate 5e-7. Dreambooth model
Training details:
850 images
learning rate 5e-7
steps 80k
batch 4
base model: AnythingV3
transforms: left right flip
No class or class images used
FAQ
Comments (10)
I was also having issues with the face, but I find it helps to keep the prompt short and sweet. and for neg prompt I'm using this (from another poster on this page):
giant, boy, man, from below, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, low quality lowres, low quality lowres low polygon 3D game, low quality lowres monochrome sketch rough graffiti, low quality lowres very ugly fat obesity scar, low quality lowres chibi, low quality lowres poorly drawn bad anatomy, low quality lowres graffiti unbecoming colorfully, low quality lowres incoherent background, low quality lowres long body, low quality lowres duplicate comparison, low quality lowres low quality lowres sketch retro_artstyle doujinshi, low quality lowres sketch, low quality lowres text font ui error missing digit blurry, low quality lowres JPEG artifacts signature hazy bleary, low quality lowres monochrome parody meme, low quality lowres historical picture, low quality lowres disfigured mutated malformed twisted human body, low quality lowres futanari, low quality lowres tentacle skeleton, low quality lowres vomit suicide death dirty, low quality, bad art, bad anatomy, asymmetrical, censored, bad art, glasses, disfigured, no arms, extra arms, extra fingers, multiple arms, more than two legs, blur, blurry deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra_limb, ugly, poorly drawn hands, text,
Mind sharing the data set photos ?
You can get similar results with less than 300 hand picked images from the web. Just pick ones that are decent quality and not to messy. No labeling needed, and no resizing needed with Kohya's Dreambooth/LoRA trainers. Just exclude any smaller sized butts.
Will you make a LoRA version of this model?
Yes, here you go!