Final epoch, posting earlier epochs later as they are less likely to cook the image when combined with other higher-strength loras





Renders models seated with their legs cross (trigger words: sitting, crossed legs). Trained on anime styles. Seems to work well with other Loras given a bit of juggling in the strengths.
Lower epochs provided in other versions as they might be better for getting the pose without affecting the style as much.
I tried using photorealistic checkpoints too but the lora unfortunately leans it towards cartoonish. Might still be enough to swing generation for poses at low weights and have the image redone in img2img.
Either way, it's been about 50/50 with getting the legs to stay properly separated. I had iffy results when trying to use inpainting to 'unweld' the legs in pictures that were good everywhere except the legs.
Final epoch, posting earlier epochs later as they are less likely to cook the image when combined with other higher-strength loras
1. 轉载模型僅供學習與交流分享,其版權及最終解释權归原作者。
2. 模型原作者如需認領模型,請通過官方渠道联系海藝AI工作人員進行認證。我們致力於保護每一位創作者的權益。 點擊去認領
