260628: Updated to Krea 2. V0.5 is the first version under training, but it still performs better than I thought, so here it is. Stable version should be released days later. 升级到了Krea 2，V0.5版本是正在训练中的第一个版本，它表现得比我想象的更好，所以分享出来。稳定版应该会在数日后发布。

260228: Updated to V2, used high resolution data for training. 升级至V2, 使用了高分辨率数据训练。

260224: Updated to V1.1 file, not a big update. 已经升级至V1.1版本文件，不是大更新。

This is a Z-image lora version of Barphot, trained with over 10k images.

Overall this model can significantly improve the aesthetics of the foot and add richness to the human poses associated with feet.

In my tests, barphotZ basically doesn't affect the performance of the base model, so you can mix barphotZ with other lora models.

Sample steps should not below 8. Recommend using on Zimage Base with official distill lora or Dark Beast.

Official distill lora: https://huggingface.co/alibaba-pai/Z-Image-Fun-Lora-Distill

这个模型是Barphot的Zimage版本。

总的来说这个模型可以显著提高脚的美观性，并且增加与脚相关的人体姿势的丰富性。

在我的测试中，barphotZ 基本不会影响底模的性能，所以你可以将BarphotZ和其它lora模型混合使用。

采样步数不应低于8. 建议在Zimage Base上使用官方Distill LoRa或Dark Beast。

官方Distill LoRa：https://huggingface.co/alibaba-pai/Z-Image-Fun-Lora-Distill