Hi! this is the last of 4 LORA's i trained based on the PastelMix Model by andite.
It tries to replicate the style without taking too much from the original model.
I recomend using:
Normal Version at 0.55 - 0.7 Weight
Minified version at 1.0 Weight
Each one was trained in a Different dataset so i cant say wich one is a better version.
This Version was trained on Abyss Orange Mix 2 - Hard (expect compatibility with NSFW)
All samples were generated at 320x320 Hires Fix x2 Latent Clip skip 2
Description
FAQ
Comments (2)
You know you can create multiple versions of the same model inside single url/entity, right? Like here https://civitai.com/models/6142/dolls - v1 and v2 both in the same url/model
Yes, but since they are all different datasets and different ranks etc, i found out with my SamDoesArt embeddings that most people will just download the first one. When it comes to the ones that included NSFW in the dataset that would make all the other versions NSFW too