Using 2 custom LoCon to fine-tune Pony.
1 dataset with 7.7k images, tagged with wd-tagger-v3. All training images are filtered and rated hq by aesthetic-shadow-v2.
hq
Trained in Pony & Animagine XL 3.1.
This is the only time you will see me posting a Pony-based model. I swear to God, Pony model is stubborn AF.
'cuz finetuning Pony is a pain in the ass.
Use as you usually use Pony.
