I used low guidance (4) for my test generations.
I use a watercolor style for a lot of my stuff, and so I tested with a guidance of 4. If your results are too similar to each other, or overbaked, try lowering your cfg/guidance.
An unremarkable LoRA for an unremarkable mercenary. Made to test civitai's Illustrious trainer with a very small dataset. Slightly overfitted, especially when using the trigger word, so it's best at middle weights and if you only prompt for grey hair, animal ears, and the eyepatch. Mileage may vary with the sword.
Please post your results and leave comments so I know how to improve the model!