v1.0
CyberRealistic Pony merged with Big Lust, and a dusting of GonzaLomo.
This 'redux' version has no additional fine-tuning, and the full merge recipe workflow is included in the metadata.
Notes
For character and concept LoRA training, I have had excellent results using Big Lust v1.6 as the base model with settings:
resolution 1024
rank/alpha 128
LR 2.5e-5
cosine scheduler w/250 steps warmup
optimi lion optimizer
batch size 8
and between 100 - 1000 LLM tagged images w/masks. I use Qwen3-VL-30B-A3B-Thinking
~step 1250 is the sweet spot
My showcase images are not actually using the DMD2 LoRA, however it is present in the workflow in an inactive branch which is confusing the Save Image w/ Metadata Universal node. The model does work with DMD2 but I am not a fan. I recommend using strength 0.7, CFG 1.3, 14 steps, LCM sampler with it if that's your thing.
I have not yet fully tested this for compatibility with Pony & SDXL LoRAs.
TBH I don't like the structure or fine details produced by this model as compared with my banned versions. Some work is required to fine-tune it. I am working on undoing some of the NSFW-by-default bias coming from Big Lust by fine-tuning this model prior to merge.
Adding GonzaLomo was likely a mistake; my impression is it seems to be overbaked with compression artifacts and is heavy handed with adding watermarks to gens. The merged model then seems to fuck up hands more than my own fine-tunes. TL;DR filter and pre-process your datasets. I have tested candidates for the next version with less/no GonzaLomo and I like the results more.
Description
FAQ
Comments (10)
I've used ten's of models and tried everything, this is the best model that ever existed , thank you for making this!
I have a question psychologicau , how can I prevent the nsfw content? I tried everything and still gives me naked girls
Can you give an example of unexpected output? That will help checking which base model / merged layers might be messing it up, and help target where to fine-tune it for the next one
I tend to use a minimal negative prompt as I believe if the model needs it, then the model sucks. Using these in the negative might help to tame the Pony base:
rating_explicit, rating_suggestive, nsfw, nude, nipples, pussy, genitals, penis, cleavage, suggestive, erotic, pornographic, topless, bottomless
I'm about to try out a character LoRA on it, so I'll poke it with some of these also. But examples would also help.
@psychologicau Thanks for reply! the only unexpected output is when I make a female character, here boobs are always out, I tried everything in the negative but still not working, everything else is so accurate in both positive and negative prompts
@psychologicau Thanks for these , I will try them now and will see if they help
I have just thrown the dataset used for the https://civitai.com/models/1907905/human-focus-photography at a bigLust fine-tune, once it's done (in a couple of days) I will see if this can be mixed into the model to tame it a little
@psychologicau it would become a perfect model, thank you!
@bigdog11 well it looks like I'm being blocked from Civitai tomorrow, so might be a Tensor art release again.
Regarding the Australian Access Block
I was hit with the login popup last night: As of March 16, 2026, Civitai is blocking Australian users. While presented as a compliance necessity, it feels like a "too hard basket" reaction that leaves the local community in the lurch.
As a single parent of a young child myself, I am not blind to the risks of the internet. I’ve been vocal with MPs via public social media posts and in online petitions because I believe in education and stronger harassment laws over blanket bans and "prohibition" style restrictions. Simply delaying a young person's exposure to the internet doesn't solve the underlying problems; it just stunts their digital literacy.
However, I also believe Civitai’s response is a knee-jerk reaction in the same order as the flawed legislation itself. By citing the "Pornhub precedent," they are using a political playbook to trigger voter backlash rather than seeking "reasonable steps" for compliance—which, as a smaller US-based entity with no local assets, they are arguably better positioned to navigate than the tech giants.
It’s disappointing to see both the government and the platform choose the path of least resistance at the expense of creators and the community.