DENIZENS OF THIS WASTEBIN,
I emerge once more from the comfortable stone I've lived under for a few months now, with a brain leaking out of my ears, new pronouns, and a desperate want for estrogen.
As such, during this tumultuous time near Christmas, I decided to gift you all with the produce of my brain leaking out of my ears.
The subject of my absolute and total (God, you have no idea how much I want to be one) is Slime Girls. So, thus, I scraped and spied and plied through Pixiv and chewed up and ingested as many of my favorite artists who produce brainrot of the kind I like.
Their names are omitted, as some do go after AI generations, and if you all found out what I like, I wouldn't hear the end of it.
227 images, 16 epochs, pure concentrated brain rot.
Go for weight 1.0, which is reasonably safe, though if you look at the images I generated in testing, you can go as low as 0.8 or as high as 1.5-ISH.
Generates incredibly, incredibly vivid, sloppy, gloopy, sticky slime, girls. There's a lot of furry in the mix, and it might have a predisposition to generate comics or sequences; however, I had little time to test that assumption as sadly, I have to be a functional member of society in a spinal industry and go to work.
Enjoy, post your generations and have fun, you all.
Description
I'm modelin' different.
This batch ain't nothin' to me, man, I'm in the GPU I'm crunchin' the feature vectors in the cloud, we trainin' Hypernets Trainin' that Style Transfer, raw concept art, Photobash deluxe, CLIP interrogator, noisy, fine-tuned base.
We usin' Sh*tpost data I'm on ten epochs loadin' in a Discord-dump prompt.
We trainin' Fidelity-Loss boogers
We pumpin' that good Latent Space, Diffusion, Jordian Jibbies (the rare style)
They must have Catastrophic Forgetting, they forgot that I'm high-res.
That Overfitting effect hittin' That model output look like a DALL-E 2 mistake.
We testin' scripts in a notebook, Blowin' the Vibe Bubbles