Kissing LoRA for Wan2.1
| Property | Value |
|---|---|
| Base Model | Wan2.1 14B I2V 480p |
| Training Epochs | 30 |
| Model Format | LoRA Safetensors |
| Resolution | 480p |
What is kissing LoRA?
The Kissing LoRA is a specialized fine-tuning of the Wan2.1 14B I2V model, designed specifically for generating realistic kissing animations. This model enhancement was trained on 50 seconds of carefully curated video content, comprising 11 distinct kissing scenes, each individually captioned for optimal learning.
Implementation Details
The model utilizes the Diffusion Pipe for Training framework and integrates seamlessly with ComfyUI through a modified version of Kijai's Wan Video Wrapper workflow. The implementation requires specific trigger words ("k144ing kissing") and operates with recommended settings including a LoRA strength of 1.0, embedded guidance scale of 6.0, and flow shift of 5.0.
- Compatible with ComfyUI modified workflow
- Trained on diverse kissing scenarios
- Optimized for 480p video generation
- Includes specialized trigger phrase system
Core Capabilities
- Generation of natural kissing animations
- Consistent results across different scenarios
- Support for various environmental contexts (mountains, lakes, indoor settings)
- Flexible prompt structure for easy adaptation
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in generating realistic kissing animations through careful fine-tuning of the Wan2.1 base model, offering consistent results across various scenarios while maintaining natural movement and interaction between subjects.
Q: What are the recommended use cases?
The model is designed for creating romantic scene animations, particularly focusing on kissing interactions between two people. It works best when combined with clear environmental context and specific character descriptions in the prompts.
