⭐If you have the means, please do support the original artist's creative work! ⭐如果您有条件,请支持原创艺术家的创作!
A LoRA that reproduces the drawing style of the artist Kase Daiki (加瀬大輝)
一个目的是还原画师 Kase Daiki (加瀬大輝) 绘画风格的 LoRA。
Artist page: https://www.pixiv.net/users/273185
Trigger word: @kasedk7
- 必须写,放在提示词最前面;漏掉它这个 LoRA 基本不激活。
Training set: 175 images, 99% portrait · MOSS-Image-Base · 1000 steps · dim 32 / alpha 32
Tips
- Start at strength 1.0 — with alpha = dim = 32, 1.0 is the full value, not a midpoint. Sweep 0.8–1.0.
建议从 1.0 起,1.0 即满值(alpha=dim=32),在 0.8–1.0 之间微调。
- Trained on portrait images (99%) — vertical framing is the native range. For wide compositions, raise resolution first instead of pushing strength.
训练集 99% 是竖构图,竖图最稳;要横构图先提高分辨率,别靠加权重硬撑。
- Optional negative prompt for clean output:
lowres, bad anatomy, bad hands, watermark, text, signature, jpeg artifacts
小提示
- 触发词 @kasedk7 必填,放最前。
- 强度 1.0 是满值,别往上加。
- 竖构图原生态;横图先拉分辨率。
- 负面词可选,上面那串是通用干净版。
Note
- 4-step turbo 的采样参数(cfg 1/ 8–12 steps)是通用起点,本 LoRA 未逐档实测,按出图效果自行微调。

