<p>A LoRA trained on 400+ images by the artist Geonjeonji. No trigger word needed. All sample images were generated with strength 1.</p><p>This was my first  time training a pony LoRA and it honestly turned out a lot better than I expected. It has a tendency to produce black and white images so i would suggest putting "black and white, monochrome" in the negative prompt. I'd also recommend a lower ADetailer face denoising strength to preserve the style in the face. The sample images use 0.2. I might retrain this at a later date with a refined dataset but for now I want to try some other things.</p>

Files with SHA256: d05b146a87a0d7cd33dbb7acb19868fca903b3fda1df4d66ec0aa3b424837507