Recommended Base Model
I’ve had the best results using this LoRA with Tsubaki. The colors, lighting, and overall CG look stay very consistent on that model, so I highly recommend starting there.
Suggested Settings
After a lot of testing, these settings gave me the cleanest results:
Sampling Steps: 20
CFG Scale: 7
Face Fix: Off
High-Res Fix: Off
Even with low steps and no extra processing, the images come out sharp and stable.
How to Prompt It
For the best outcome, try building your prompt in this order:
-
Describe the character (appearance, expression, pose, hair).
-
Add any outfit or scene details you want.
-
Then finish with the style tags below—they help keep everything in the CG look this LoRA was trained for:
honkai_star_rail_cg_style, highly detailed, sharp details,
refined textures, cinematic lighting, soft shading,
light reflections, subtle glow, vibrant colors, rich contrast,
dynamic pose, flowing hair, motion dynamics,
intricate costume details, polished r
These tags usually give the most reliable results and keep the style consistent across different prompts.