Recommended weight: 0.9

While the LoRA can function at 1.0, lower weights tend to preserve better composition and avoid over-emphasizing facial structure or stylistic features. For most checkpoints, 0.85–0.95 provides the best balance between character accuracy and image coherence.

This LoRA was trained to recreate Gao Shiba with strong consistency across facial proportions, hairstyle, expression, and overall visual identity. It performs well across anime-focused checkpoints and is optimized for portrait shots, cinematic compositions, and high-detail character renders.

Recommended to be used with shadow shift