Pony v2Well... here it is. Don't know what happened during the training process. It turned out good in terms of actions, poses, etc, but somehow even worse with styles than a v1. Some style Loras work, but you have to crank them up to the maximum for that, as a result... it tends to overcook the image and certain details become smudged. Guess, it calles for an eventual v3.Pony v1My first Lora that turned out good enough to publish. I honestly have no idea how well it works and how flexible it is, so please provide feedback so that I can do better next time!

[Underworld] William Corvinus, the first werewolf (Pony)

Reworked the dataset. Not overtrained, more stable, more flexible in terms of actions and poses, but, sadly, less flexible in terms of style.

Files with SHA256: 8109af8ddb218a59ebcf1f7323c76854e9277000859cc18eaeed5067b6f82a53