Siofrafication - v2 Showcase

Siofrafication V2 showcase! My semi-realistic anime style LoRA, now trained on 178 images with proper booru-style captions.

What's new in V2:

• Much larger dataset (178 vs V1's smaller set)

• Better male/female character separation (even no character images were in dataset so you can create scenery)

• More consistent style across different prompts

How to use:

• Trigger word: carmV5style

• Recommended weight: 0.7-1.0

• Base model: Juicy Majic Illustrious V5 (or any Illustrious-based checkpoint)

• Use booru-style tags, not natural language prompts if possible

• Clip skip 2

Works great for both 1girl and 1boy generations!