This LoRA is designed to generate images of Jinx, capturing her distinct appearance and chaotic energy as seen in the Netflix series Arcane. Through extensive testing, I’ve found that the weight of the LoRA doesn’t drastically influence the output. Instead, the CFG (Classifier Free Guidance) setting plays a more significant role in determining the final image style. Lower CFG values produce a more oil-painted aesthetic, while higher CFG values bring the images closer to Jinx’s depiction in Arcane.

At a weight of 1.0, the images can lean towards a simplified, cartoon-like style reminiscent of the League of Legends video game or even Fortnite. Increasing the weight up to 1.6 refines the output to better align with the more detailed and stylized look seen in the show.

Despite what conventional wisdom might suggest, the model’s training parameters were set to maximize learning within the constraints of Civitai’s settings. While the LoRA might appear somewhat inflexible in generating Jinx in non-typical scenarios, this could be due to my own limitations in prompt crafting. I encourage the community to experiment and possibly achieve even better results.

Recommended Settings:

Weight: 1.2 - 1.6

CFG: 4 - 6

Tags Used: Jinx, braids, blue hair, cropped top, striped shorts, wild eyes, mischievous smile, playful, punk style, chaotic, pale skin, tattoos, boots

Waiver of Liability: By using this LoRA, you acknowledge that the creator is not responsible for how it is used. Any misuse or creation of inappropriate content using this LoRA is solely the responsibility of the user. The LoRA is provided as-is, with no warranties or guarantees regarding its functionality or suitability for any specific purpose. Please use it responsibly and in accordance with the Civitai TOS, as well as all applicable laws and guidelines.