Trained to strengthen the concept of a karate gi. Images were selected to more or less conform to the same style of outfit.Not quite where I want it to be yet. Has a tendency to get swallowed up by other parts of the prompt or other LoRAs. Might have to separate out different form of the outfit into their own activation tags at some point and rework the dataset for better consistency.Long story short, if you want to make this LoRA work you might have to adjust weights a bit (raising the karate gi weight or loweing the other model's weights or both).

Files with SHA256: 8a056805c188275a3b931b68350444a21e0cd3d04970d40eb628764326c3327a