<a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/355065">Asanagi</a> Style Lora.*could act in a funny overtrained way, kill its strength to maybe rein it in a little bit**impossible not to have a big chunk of the dataset be monochrome/black and white, so often it outputs that. Consider putting "monochrome, black and white" in the negative. Maybe even give it more strength like 1.2 or 1.3

Asanagi Style [IL] | [PonyXL]

Example images with <a target="_blank" rel="ugc" href="https://civitai.com/models/415952/pulenkompot-ponyxl-checkpoint-mixes">this</a>.

PulenKompot-Asanagi Style-PonyXL.safetensors

PonyXL

<a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/355065">Asanagi</a> Style Lora trained on/for PonyXL and all its merges.*could act in a funny overtrained way, kill its strength to maybe rein it in a little bit

Files with SHA256: b6faced8858c091653ab9620023d51a09c970cb662522c9f20f91c96dac318f3