v1.0 Release!
LoRA was trained on, and all images were generated with, Counterfeit v2.5
All examples use the seed of "1" with a batch size of 8, meaning no seed cherry picking was done.
Description
v0.3
Trained on 23 character portraits from Persona 5 Royal and Strikers, 300 per image, batch size 2 with corneos7thHeavenMix_100.
All example images generated at CFG Scale:~2 at 512x512 with LoRA strength at ~0.7 and Clip Skip 2. All examples except Baseball Cap Girl were generated with public LoRAs/Hypernetworks of the characters.
"1girl/boy, portrait, solo, white background" is how I've started most of my prompts, but I can't guarantee that's the best way to go, feel free to experiment yourself.
FAQ
Comments (7)
This looks amazing, dude, thank you so much for making this. I really want to learn how to make style's like this, can you explain how you did it? Like is there a guide?
I followed this guide: https://youtu.be/k5imq01uvUY
I honestly have very little idea what I'm doing, most of this came from trial and error. I just have a killer PC that is able to waste time doing stupid things in the hopes that something good happens.
@todokete Thanks! :) posted a review as well
Hey where did you get the other Lora? Like Zelda or Hex Maniac?
The Zelda one was from here: https://civitai.com/models/4959/princess-zelda-lora
I got the Hex Maniac, and few others, from: https://aids.miraheze.org/wiki/LoRAs
@Todokete interesting, i couldnt get any of the lora's fromo aids to work :sadge:
@alfajoebed Strange, I suppose I did rename the files to the character names, like "HexManiac" so I'd know which was which. After that I just selected the LoRA and added the character tag and a few descriptors.