This LoRA is trained on images of the bulky Sarkaz mercenary NPCs from Arknights, particularly the Sarkaz Centurions.

This works reasonably well (given the meager training dataset) but isn't the most stable. While the trigger word, sarkaz mercenary, works fine on its own, additional prompting may be required to get details down, especially for complex prompts or if multiple LoRAs are involved. Consider adding mask, armor and horns or other clothing details.

Weight should be fine around 1, but if the images become washed out or fried, consider lowering it to ~0.7 (or lower if needed, although results may become less stable).