Miratsu

Embark on a delightful journey into the charming world of miratsu's unique art style!

This LoRA is meticulously trained to capture the distinctive chibi (Q-version) anime style of the celebrated artist miratsu (X/Twitter | Lit.link). Known for their adorable fan art of beloved characters like Hatsune Miku, miratsu's signature aesthetic is now available for your Stable Diffusion creations.

Key Features & Style Highlights:

• Distinctive Chibi Style: Generates characters with exaggerated features, large eyes, small noses, and round faces, embodying the quintessential chibi charm.

• Pastel & Dreamy Color Palette: Specializes in pastel colors (e.g., pastel pink, pastel purple, pastel blue) and soft shading, giving your creations a gentle, ethereal glow.

• Kawaii & Expressive: Excels at rendering cute expressions, blush, and kawaii poses, bringing out the inherent cuteness of the characters.

• Signature Details: Often incorporates black outline and simple background or white background elements, consistent with miratsu's original works.

• Versatile for Characters: While drawing inspiration from miratsu's fan art, this LoRA can be used to generate your own original characters (OCs) or existing IP in this delightful style.

How to Use:

1. Base Model: This LoRA is optimized for Illustrious and its derivative models. Using other base models might yield different results, but experimentation is encouraged!

2. Trigger Word: Always use miratsu in your positive prompt to activate the style.

3. LoRA Weight: A weight between 0.8 and 1.0 is generally recommended for optimal style transfer. Adjust to your preference.

4. Recommended Prompts:

   • To enhance the style, include tags like:
     chibi, anime style, digital art, pastel colors, kawaii art, cute expression, exaggerated features, large eyes, soft shading, black outline, simple background, white background.

   • For character specifics, add details like:
     purple hair, blue eyes, pink bunny hair clip, cat ears, white shirt, black vest, etc.

5. Sampling Settings:

   • Resolution: Samples were generated at 1024x1024. We recommend generating at this resolution or higher for best quality and detail.

   • Sampler: DPM++ 2M Karras was used for sample generation.

   • Negative Prompts: Use standard negative prompts to avoid common quality issues (e.g., lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry).