Flaaffy LoRA trained on NovelAI with Kohya training scripts using the DAdaptAdam optimizer at 512x512 resolution. Skin coloration is very sensitive to any colors in your prompt, I got completely blue skin when prompting for (lacy blue bra for example. I had to add (tail orb) to my prompt to get a more consistent tail. Needed to add (hooves) to get proper feet whenever they were in frame. Should have wide compatibility with model mixes based on NovelAI, which is most of them. Trained on a mixed male/female dataset.
Description
Initial release
FAQ
Comments (4)
Awesome, It'd be even more awesome if you'd be able to do all of the Pokémon one day,
I wish dude, I wish. I'm gonna have to get creative once I start getting to the mons with <30 good source images to train with 😭
Slight misspelling, it is Flaaffy. Not Flaafy.
Oh shit, I misspelled it through the entire stack, including the LoRA filename 💀. I've corrected the spelling on here and I'm going to rebake this LoRA to get the filename (and name in Automatic1111) to be spelled right. Thanks for letting me know!