English

A style slider for MOSS-Image-SFT that moves images between modern anime illustration and realistic photography. Negative weights → anime; positive weights → photography; 0 → base-model output. Start at -1.0 or +1.0 and adjust to taste. No trigger words are required.

Recommended sampling

• Sampler: er_sde · Scheduler: simple

• Steps: 25 · CFG: 3 · Resolution: 1024 × 1024

• SLG: uncond skip, layer 26 (0-indexed), start 0.0, end 1.0

• Apply LoRA to the model only; CLIP strength 0. Negative prompt: empty.

Feel free to leave comments and share photos.

中文

适用于 MOSS-Image-SFT 的画风滑块，可在现代日式动漫插画与真人摄影质感之间调整。负数偏动漫，正数偏摄影，0 为底模原始效果。建议从 -1.0 或 +1.0 开始，按喜好调整，无需触发词。

推荐采样设置

• 采样器：er_sde；调度器：simple

• 步数：25；CFG：3；分辨率：1024 × 1024

• SLG：uncond skip，第 26 层（从 0 计数），起止区间 0.0–1.0

• LoRA 仅作用于模型，CLIP 强度为 0；负面提示词留空。

欢迎多多留言，晒图。