English Description

Experimental LoRA based on Safir from Snowbreak: Containment Zone.
This model was mainly trained for testing purposes, including parameter experiments and character feature extraction.
The dataset and tagging process were handled rather casually, so consistency and accuracy are not guaranteed.

It may perform surprisingly well on some checkpoints, while showing noticeable drift on others.
Best used for experimentation, tweaking, or further fine-tuning.

Feel free to adjust weight, CFG, and prompt combinations to achieve better results.