Train base on Pony + normal SDXL checkpoint. with 600+ images from overwatch, sort a little from my owstyle Lora.
Aiming for making CGI faces with basic prompts.
Used it like how you would use Pony.
Post images are made with basic prompts like:
Prompt: score_9, score_8_up, score_7_up,score_6_up, score_5_up, highly detailed,1girl,
Negative prompt: score_1, score_2, score_3, score_4, source_furry, source_cartoon,
Some used Adetailer to fix the face
The best way to sampling I find is Euler A
Trying to make sdxl work like sd1.5 with just a few basic prompts. The way SDXl works is so dumb and numerous. Glad pony is here to save it.
Give it a try, it is my first Trained checkpoint so not sure if it is working right :-/
Share some posts :D
If there is something wrong let me know here or on my discord :)