About this version
An experimental Lora you may use in your merges
It likes to mess up faces and was trained with Virile Reality V3 BETA 3 as a base
it uses Clip Skip 1
In Version 2.0 I now fixed having it send the person into the forest all the time, and legs also now look beefier.
maybe try it at like 70% strength, and something to fix faces
To get best results you need to use the VAE vae-ft-mse-840000-ema-pruned
Description
New Version, less forests, and bigger legs,
You should start with a sentence, then throw in additional triggers
This Version features bigger legs and less forests, and slightly improved look of flaccid heads
FAQ
Comments (6)
Since this was a concept that could be applied to any model, In my opinion, you should have trained on the 1.5 base model. Using a specific model is good if you want super specific thing with fidelity to details that only a specific model provides, like face details and realism.
Virile Reality is a model with too much distortions. I haven't tested it, but it probably makes your Lora not very flexible. But I could be wrong.
Anyway, Thanks!
The base model didn't have enough penis knowledge to properly work
@encante0123 but you don't need the base model after it's finished training. You can use the lora with a model that knows about penis after it's done. Or use a penis lora.
@diogod ohhhh i see, so train it once with the main one, then do it again with whatever
@encante0123 to clarify. You should train a Lora using the base 1.5 original model. Then, after the lora is ready for usage, you can use it with whatever checkpoint model you want. Doing that ensures flexibility since all the fine tuned models (like reallisticvision, etc) are derivatives from the original 1.5 model. So basically, unless you are training anime or a face, you should use the original 1.5.
Wish it would come in Pony XL.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.