💡 Core Style This LoRA is designed to generate high-quality Japanese anime-style images with soft lighting and a delicate atmosphere. It reliably produces: Delicate facial features and expressions. High-quality, soft, and warm-toned lighting and shading. Clean and professional animation line work.

v1.2 Update:

Outfit and Detail Enhancements

Simplified Elegance: Transitioned from the heavy lace mandarin collar (v1.1) to a more casual, sweet-style blouse with a bow knot.

Accessory Addition: Incorporated a leather belt to add structure and detail, enhancing the character's overall completeness.

Composition and Lighting Refinements

Softened Lighting: Reduced the intensity of the warm sunset light, resulting in a more evenly lit scene and better emphasis on the grass texture and depth of field.

Refined Facial Expression: The expression is standardized to be sweet and clear, reducing the strong blush/coyness seen in v1.1.

If you want to try this LoRA in DiT Model, You can check this link: https://pixai.art/model/1940734286665471614-Soft-Color-DiT

🔑 Key Usage Guide

Stable version is v1.2
Trigger Word (Required) Ensure this word is placed at the beginning of your prompt to activate the LoRA effect. Trigger Word: Soft_Color_XL, warm lighting, soft shading
Recommended Settings Model: Haruka v2 LoRA Strength: 0.7 ~ 1.0 Sampling Method: Euler A CFG Scale: 3 ~ 5 Steps: Determined by Prompt Complexity
Recommended Prompts To ensure image quality, it is advised to always include these quality tags: masterpiece, best quality, highly detailed, soft lighting, warm lighting, delicate shading, beautiful eyes

⚠️ Warning: If the base model you are using tends to produce images with excessively high contrast, please lower the Strength of this LoRA (e.g., reduce it to 0.8 or 0.7) to prevent the image from becoming overly sharp or distorted.

This is my first time training a LoRA model. Please let me know if you have any suggestions or questions!