ANIMA

白银之城风格，包含2d与3d，由于我炼的不算好，所以角色服饰可能会带有不符合的装饰。

Silver City style, supports both 2D and 3D. Training quality is limited, so characters may have inconsistent accessories on clothing.

默认2d风格，在提示词增加3d可以让风格变立体（建模风格）。

Default: 2D art. Insert "3d" into prompts for textured model-style rendering.

因为数据集是从别人的视频截图下来的（我没测试资格我真的要生气了:( ）,所以默认出图大概率是半身，需要比较细致的提示词去控制人物姿势。

Training data sourced from video screenshots. I lack game test access, which is frustrating :( Images default to shots; use detailed prompts to adjust character poses.

这个lora炼了五次，第一次有点效果之后用了大量时间去完善数据集想效果更好点，没想到越高越糟糕，炼丹真是个玄学。

Illustrious-3d

推荐开启二次采样。

Recommended to generate images with sampler two‑pass.（idk how to translate this sentence）

和anima版本还是有点区别的（例如眼睛和质感），不过我认为总体比anima好一点，下一个该优化anima版本了。

There are still some differences compared to the Anima version (e.g. eyes and material texture). Overall though, I think it’s slightly better than Anima. Next up, I’ll work on optimizing the Anima version.

如果你觉得风格太弱了，拉高一点权重（1.1-1.2）

If you feel the style is too weak, increase the weight slightly (1.1‑1.2).

Illustrious-2d

很遗憾，这是个半成品。在未来我会更新优化。

Unfortunately, this is a semi-finished product. I will update and optimize in the future.

比起游戏画风，这更像是主美小纪自己的画风，并且一些角色使用这个lora会是3d效果，我想不通为什么没法用提示词控制，可能是我的数据集/打标问题。

Compared to the game's art style, this is more like the protagonist's own art style, and some characters using this lora will have 3D effects. I can't figure out why I can't control it with prompt words, maybe it's a problem with my dataset/marking.