<p>A LORA for the Blazblue character Bullet, trained PonyXL V6. Trained at 1024x1024 at 32 Dim for 10 epochs totaling 15000 steps, this character LORA was partially an experiment in what effect higher repeats per epoch would have on character LORAs. It seems to be better at capturing minute details of the character, such as her steel-toed knee high boots and hanging belts, but still struggles with her gauntlets and might be a little overfit. Given the detail and complexity as well as the uniqueness of her weaponry, I consider this to be an innate limitation of current AI image generation technology. If you find them ugly, I recommend adding "weapon" or "gauntlet" in the negatives.<br /><br />As this LORA was trained on a variety of official and on-model fan art of both the SFW and NSFW variety, there should be a high degree of flexibility for what you can prompt. </p>

Files with SHA256: 3b09cef9a02fd7cf8131f60b02f9fcc8403d1c01cb2ac34314bda3138dc9b0a2