sexual situations

physical violence

disturbing

male nudity

hanging

hate symbols

nazi party

revealing clothes

weapon violence

female swimwear or underwear

male swimwear or underwear

partial nudity

white supremacy

adult toys

graphic male nudity

illustrated explicit nudity

nudity

pg-13

convenient censoring

oral invitation

exposed female nipple

female nudity

undressed

male underwear

female swimwear

genitals

female underwear

thick thighs

breasts out

strapless leotard

vore

breast out

one breast out

huge breasts

gigantic breasts

huge butt

covered nipples

hair over breasts

no panties

sitting on face

anal

dildo riding

downblouse

oral

porn

futanari

hentai

nude

lingerie

sexy

nsfw

child on child

incest

graphic violence or gore

wide hips

emaciated bodies

sexual activity

sexual intent

suggestive

self injury

extremist

hate speech

diapers

urine

scat

graphic female nudity

corpses

peeing

blowjob

latex clothing

swimwear

bukkake

fellatio

cumshot

implied fellatio

eat_cum

cumdrip

cum in pussy

cum on face

after fellatio

cum on hair

cum on body

cum on tongue

cum on hands

cum in mouth

triple fellatio

autofellatio

fucked silly

cum on pussy

pov fellatio

Trained using OneTrainer on an RTX 4070 SUPER 12gb, using 200+ illustrations by sushi yuusha toro from black souls 2 as well as the bonus contentUse the sushi yuusha toro tag for the style. (also "sketch" might help, as well as removing some negative/quality tags)All the images were auto tagged by <a target="_blank" rel="ugc" href="https://huggingface.co/p1atdev/wd-swinv2-tagger-v3-hf">wd-swinv2-tagger-v3-hf</a>

experimental flux version, trained on-site

flux-1.0

trained for illustrious 0.1, no idea if anyone even uses it anymore

illustrious-0.1

a wai nsfw illustrious version, as requested

wai-nsfw-illustrious

using a logit normal timestep distribution, should stick closer to the style

vpred-1.0-v4

apologies for another update, i'm testing timestep shift with a value of 2consider giving feedback on which version works best for you

vpred-1.0-v3

once again using prodigy, now with DoRa enabled and debiased loss estimation it comes pretty close to the style, still looks a lot aesthetically better than the original images, without overfitting on the dataset

vpred-1.0-v2

a test version using the prodigy optimizer and text augmentation, constant loss and DoRa disabledbatch size 4, 2 image variations (1 repeat) 4 text variations, 10 epochs, might be a bit underbaked

vpred-1.0-prodigy

epspred-1.1

vpred-1.0

[NoobAI | Illustrious | Flux] black souls / sushi yuusha toro style

00484-2085447829.png

Type	LoRA
Stats	69 3 18
Reviews	Positive (13)
Published	Apr 15, 2025
Base Model	Flux.1 D
Training	Epochs: 10
Hash	AutoV2 967C6BA5BA