it is not a normal ceckpoint-model, its for fine tuning your pics !
not new trained only a merge of pix2pix and realv30
my first try of a new PIX2PIX model with better quality and may NSFW content
Please tell me what you do with such a model?
save it in default model folder
use it like:
https://stable-diffusion-art.com/instruct-pix2pix/
main idea is to work with img2img you can write a prompt like: change A into B, and play with CFG Scale and Image CFG Scale careful.
works also with inpaint little better than normal inpaint models but only with change colors etc. If you have a png image with areas erased (remove background) and mask it, everything will always be changed no matter what the settings, in my model more the masked areas would be changed.
dont use BadDream, UnrealisticDream.
-> if inpaint use only (original) and you will see an extra button
Description
FAQ
Comments (8)
pruned version plz
plz ;)
这 不是和CON1.1 IP2P重复的了吗？
english pls? ;)
不，模型只是原始pix2pix和realV30的组合。
No, the model is just a combination of the original pix2pix and realV30.
Hey buddy,
Its very impressive.
Could you please tell us what weight did you use for model merging between two models?
thy
but sry, no i cant i forgot ^^ , it was not that complicate ...
RuntimeError: Given groups=1, weight of size [320, 8, 3, 3], expected input[2, 9, 96, 64] to have 8 channels, but got 9 channels instead
it woks in automatic1111, it needs (instruct-pix2pix.yaml) but is standard, and only for images aprox 740x740