Please use on https://github.com/AUTOMATIC1111/stable-diffusion-webui
Base Model
Chilloutmix/chilloutmix_NiPrunedFp16Fix.safetensors [59ffe2243a]
512 * 768
Tested On
Chilloutmix/Chilloutmix.safetensors [a757fe8b3d]
AdamW optimizer, trained on RTX4090, my inital try to use LoRA.
Using Latex experments images. (Oh! I love there shoots!)
Still going to updating!
TODO:
Add pic from refect desire (tested, interfer with current trained pic, will open another repo)
More pic
style focused
Additional picture source is welcomed.
Description
1000 images, mainly alex and juli
The output is better, watermark need to solve
Let us make it a try!
I reviewed the training process, it is likely overfitting. Concidering how similar the face to alex.
FAQ
Comments (6)
Thats Larum.... latexperiment .... yeah the watermark for training is terrible with that set.. have 120GB of his work but did not actually get to his sets yer.
Thanks testing it right away with our models so far
🤟 🥃
watermark is able to solve, by adding reflect image, or simple crop the image. currently I don't have enough time, i am working on my paper. The diversity of the image is rather difficult to solve, since there are only a few models.
Thanks for your comment, currently i only use image from 2010-2013.
The cover image is uncannily like Alexandra P-tt-r, do you think you could replace it with a fictional person instead?
I will do it later
Runs very well together with Hentai Diffusion @ 0.2 weight.
what portion of the dataset is Alexandra Potter?