Realistic AI Parody - SDXL Getting Started
Basics
- Have a computer with a graphics card. Preferably Nvidia - AMD can work, but will be slower. The faster the better, obvs, but anything from the last 10 years with at least 8GB VRAM should work.
- Download and install a Stable Diffusion UI. Forge or Fooocus are good for beginners. ComfyUI is more advanced. Stability Matrix is a "Multi-Platform Package Manager and Inference UI for AI Image Generation" which basically makes it pretty painless to try and use different UIs on windows mac and linux.
- Download Big Lust v1.6 and put it in the models subfolder.
- Download DMD2 and put it in the LoRAs subfolder.
nb: DMD2 improves quality and speed of generating, at the cost of variety and prompt adherence. You must use it with certain settings (see below), and negative prompts do not work with it. - Download subject person LoRAs from CivArchive and elsewhere. Ones trained on Big Lust work best, but any trained on SDXL will work to some extent. See below for a good list of creators to follow.
- In SD, select Big Lust checkpoint, and your DMD2 and subject person LoRA, then set the following settings:
- Steps: 7
- CFG: 1.0
- Sampler: lcm
- Type in what you want to see and press 'Generate'.
If you get stuck, or need help, or your gens look weird, come back to the thread and post your settings. Someone might help you.
For Best Results
- Try and stick to the optimal SDXL resolutions.
- Use the bigASP tags list.
- Some LoRAs don't work well without their keywords. DMD2 doesn't have one, but most subject person LoRAs do. Check 'trigger words' on the page you got the lora from or using Xypher7's lora metadata viewer, and use them in your prompts.
- Detail on faces can be dodgy sometimes, especially on wide shots. Learn to use inpainting, ADetailer, or FaceDetailer (depending on your choice of UI) to improve them.
Advanced
No LoRA available for the person you want to generate? Try training one yourself! The following resources should help you get started:
- LoRA Training Overview
- LoRA Training on Runpod for those with less-powerful GPUs
- Gathering and Captioning Training Images arguably the hardest, but most important, part of the process

Celeb LoRA creators for BigLust
BertyDastid
cptdan
__deep
detdet
diaryph
DRZ3000
enzino
GoldenSacks
jc7890
MKUltra2025
mrmustard
peanutbutter_
rustyshackleford69420
SxRisen
ThatsARepost
toxicmuffin
Also keep an eye on:
aihotgirls (Jordyn Jones)
MachineMinded
MrME (JLaw)
Unshackled_AI (trains on lustify)
Yat99 - adds/publishes LORAs made by anon
others found on civarchive:
AIENGI
BertyDastid
dollrama126
heydixmcgee
Hikarias
images101
kamikazey
leisure_suit_larry
MacDonal
MakeThemComeAliveAIArt
Makimark
OmoshiroiHS2
ParanoidAmerican
Peli86 - mostly porn girls
razzz
sassytwinklol
Signalytix
steffangund
TheGoblinQueen
TheMainManFks
loradude - not great imo, do your own testing if needed
malcolmrey
Sharing is Caring gofile - tons of loras and other resources, should be linked in OP
Sharing is Caring SDXL Celeb LORAs folder
non-SDXL LoRAs
https://huggingface.co/malcolmrey
https://civitai.com/user/malcolmrey/articles - how he do
https://civarchive.com/users/malcolmrey