LoRA of Nekokoyoshi (喵小吉)

Only her face and upper body are included in the training data.

Hand-picked 104 images for training.

Work well with 2.5D or realistic checkpoint.

Best weight 0.7-0.9.