Daioo这个作者的画风我很喜欢，于是炼化了这个lora，依然推荐配合RDBT底模使用

采样器推荐: er_sde, Euler, Euler a

触发词：@daioo

在训练这个作者的过程中，我发现总是无法完美复刻daioo这个作者那种丰富线条的感觉，所以这个画风也许只有90%的还原度？训练anima lora对画风线条之类的学习似乎不像sdxl那样简单，如果你们有能改善这个lora线条学习的方法和训练参数，麻烦请告诉我，必将活用于下一次。

I really like Daioo’s art style, so I created this LoRA. I still recommend using this LoRA together with the RDBT 0.39b checkpoint for the best results.

Recommended samplers: er_sde, Euler, Euler a
Trigger word: @daioo

While training this LoRA based on Daioo’s style, I found that I still couldn’t perfectly reproduce the rich, detailed linework that makes Daioo’s artwork so distinctive. So I’d estimate that this LoRA achieves around 90% fidelity to the original style.

Training an Anima LoRA for things like linework and artistic style seems to be considerably more difficult than doing the same with SDXL. If anyone has any methods or training settings that could help improve the LoRA’s ability to learn the linework, I’d really appreciate your advice.