Trained on Hassaku, PixAI site, 11 images. It would seem to work great. It's compatible with models like Koji v2.0 - v2.1, Dark Sushi, Galena REDUX, GrahamAingPikas Clear VAE2.3, etc. I used ze default 0.7 LoRa strength, which works well enough. You may experiment if you feel like it. I used CFG Scale "7-8", again, feel free to experiment. I'm using strong LoRa, "Add More Details - Detail Enhancer" ( <a target="_blank" rel="ugc" href="https://pixai.art/model/1622983818871110869">https://pixai.art/model/1622983818871110869</a>) for ze background as a default now, I'm not sure how good is this LoRa at generating rich sceneries or props by itself. Thumbnail model: Koji v2.1

Miyabi Yuugaoze - 夕顔瀬雅 / Maou Gakuen no Hangyakusha - 魔王学園の反逆者

Again, I don't know what SD is PixAI using, so I leave "Other"

Files with SHA256: 3881f39bd684a47ddab23973eff554f10760aa147893129982f7bff9621d76ad