📘 LoRA Overview

Experimental version has bentstand-front related content and optimized the issue of overly rigid leg movements,but stability is not fully guaranteed, so use with caution.

The training images for this LoRA were constructed using ControlNet guidance, following the composition and pose shown in these two examples, and also incorporating some elements from Part 1.

Added a front view and fixed the issue where both arms wouldn't hang down naturally. It really means a lot to me.

📌 Recommended Settings

This LoRA works best when used with expression and scene-related prompts. The training images might be a bit simple in those areas, but compared to Part 1, the expressions and backgrounds are much more flexible.

Use bentstand-front or bentstand-behind as trigger words.

You don’t need to include 1girl and 1boy in the prompt.

There are four main composition types. Below are the core prompt structures for each:

• Front View: bentstand-behind, dangling legs, dangling arms, front view

• Side View: bentstand-behind, dangling legs, dangling arms, from_side

• Arm Held Back: bentstand-behind, dangling legs, arm_held_back, dutch angle

• from Front: bentstand-front, , suspended, from front, torso_grab (unstable)

Check the sample images for more info.