thick thighs

sexual situations

physical violence

disturbing

extremist

hanging

hate symbols

nazi party

breasts out

strapless leotard

revealing clothes

weapon violence

female swimwear or underwear

male swimwear or underwear

partial nudity

self injury

white supremacy

adult toys

graphic male nudity

illustrated explicit nudity

nudity

sexual activity

graphic violence or gore

graphic female nudity

vore

bestiality

corpses

breast out

one breast out

huge breasts

gigantic breasts

huge butt

wide hips

covered nipples

hair over breasts

no panties

sitting on face

convenient censoring

anal

dildo riding

downblouse

peeing

oral

oral invitation

emaciated bodies

hentai

porn

bukkake

cumshot

cumdrip

cum in pussy

cum on face

cum on hair

cum on body

cum on tongue

cum on hands

cum in mouth

cum on pussy

eat_cum

fellatio

implied fellatio

after fellatio

triple fellatio

autofellatio

pov fellatio

fucked silly

presenting ass

presenting pussy

spreading legs

spread legs

swimwear

lingerie

bikini

latex clothing

TRPCError

Pony Diffusion has a bad habit of generating very blurry/grainy background if you dont specify them in detail or use the right negatives, so I decided to create a background detailing lora that can improve backgrounds and scenery without messing with the subject too much. Other detailing loras already exist, but most of them focus on detailing the whole picture rather than just the background.

This version of the model should retain the background enhancing ability of previous models without effecting the overall style of generations too much. It still doesnt need any trigger words and seems to work best with a strength around 0.4-0.6, but you can go as high as 1.0 and as low as 0.2 if needs be. Some issues I have noticed is less consistent crowd generation, some difficulties with animals in outdoors settings, and increased detail in 3d generations, which may mess with some styles. The model's ability to improve backgrounds of images with humans in them and images with just background may have improved slightly, however.

Pony Beta V3

Since some reviewers recommended improving crowd scenes I added some crowds to the dataset and I'm pleased with the result. If you feel the results are too realistic style, use "source_photorealisitc" in the negatives.

Pony Beta V2

Beta version. Good at detailing backgrounds, but also does slightly effect the subject. I recommend weights around 0.6 to keep the subject mostly intact, but you can raise them if you want. Shouldnt require any trigger words.

Pony Beta V1

A very alpha version of my detailer for illustrious trained on the same data set and tags as the pony version on Civitai's trainer. I am still trying to figure out illsutrious' training paramters, so this version cant even be considered a beta, but I figured I would post it just for people to test out while I figure out how to get it to work. It seems very inconsistent in providing detail and actually seems to remove detail from outrdoor backgrounds at strengths over 0.6. For this reason I recommend using it at a strength of around 0.4. The model does seem to improve prompt adherence for some reason as well for some generations, and I'm not sure why. There arent any keywords for this version, though you can try putting "blurry,blurry background" in the negatives if you have issues.

Illustrious Alpha

Background/Scenery Detailer (Pony and Illustrious)

Type	LoRA
Stats	645 87,139
Reviews	Very Positive (74)
Published	Aug 16, 2024
Base Model	Pony
Training	Steps: 224 Epochs: 4
Usage Tips	Clip Skip: 2 Strength: 0.5
Hash	AutoV2 DF9E5D5356