This is a custom-made LoRA I created. The training images consist of 45 in-game screenshots with transparent backgrounds. When using this LoRA, make sure to include more detailed background prompts. Depending on the model, the recommended weight ranges from 0.6 to 0.9—start testing from the lower end.
Known issue: The white puffy shorts are sometimes interpreted by the AI as a skirt. If the generated image does not show any white clothing as intended, you can try increasing the weight of skirt-related prompts to 1.2
