Wan-LoRA-Arcane-Jinx-v1
| Property | Value |
|---|---|
| Author | Cseti |
| Model Type | LoRA |
| Training Data | 33 images + 35 videos |
| Model URL | Hugging Face |
What is Wan-LoRA-Arcane-Jinx-v1?
Wan-LoRA-Arcane-Jinx-v1 is a specialized LoRA model designed to generate high-quality images of the Jinx character from Arcane. The model employs a unique two-phase training approach, combining both image and video data to achieve superior results in character generation and style consistency.
Implementation Details
The model utilizes a sophisticated training methodology with the following specifications: Learning rate of 2e-5, AdamW optimizer, 15 epochs for image training followed by 7 epochs for video training, totaling 5070 steps. The implementation uses a rank of 128 and batch size of 2, trained using the Diffusion-Pipe repository.
- Dual-phase training approach (image + video)
- High rank (128) for detailed feature preservation
- Optimized hyperparameters for character consistency
- Compatible with ComfyUI native nodes and Kijai's wrapper nodes
Core Capabilities
- Accurate character generation with specific trigger words
- Diverse scene composition support
- Cyberpunk and urban environment integration
- Flexible style adjustment through prompt engineering
Frequently Asked Questions
Q: What makes this model unique?
The model's distinctive feature is its dual-phase training approach, combining both image and video data to create more consistent and accurate character representations. The high rank of 128 ensures detailed feature preservation.
Q: What are the recommended use cases?
The model excels at generating Jinx character images in various settings, particularly urban and cyberpunk environments. It's best used with trigger words 'csetiarcane', 'Nfj1nx', and 'blue hair' at a strength of 0.9-1.0.
