Wan-LoRA-Arcane-Jinx-v2

Latest version of my Wan Jinx lora.

Differences:

• It is a bit weaker,
• follows the prompt better,
• Less bleeding to the other characters

Prompt

csetiarcane. Nfj1nx lounges on a hammock strung between two trees, her long blue hair spilling around her like a waterfall. She’s dressed in a simple, comfy summer dress, her bare feet propped up against the hammock’s edge. She looks at the camera with a sweet, relaxed smile, one hand resting behind her head as the other holds a cold drink with a small umbrella. The surrounding garden is lush with greenery, and the light creates a warm, inviting glow around her, enhancing the casual and carefree atmosphere

Prompt

csetiarcane. Nfj1nx and her partner share a tender moment in a secluded garden, surrounded by vibrant blooms and the gentle hum of nature. She wears a flowing dress, the fabric swaying gently with the breeze, and her long blue hair frames her face. They are seated on a bench, and Nfj1nx leans her head on their shoulder, her eyes closed in contentment. Her partner brushes a lock of her hair behind her ear, their fingers lingering for a moment longer than necessary. The air between them is thick with unspoken affection, the gentle touch speaking volumes about their bond.csetiarcane.

Prompt

csetiarcane. Nfj1nx stands in the middle of a room of a teenager, bathed in warm, soft light that accentuates the deep V-neck of her transparent white nightgown. The nightgown flows loosely revealing her naked body and shapely breasts while she runs her hand over her breasts. Her long blue hair is styled in loose waves, framing her face and adding to the relaxed sexuality of the shot. She glances over her shoulder toward the camera, a playful, confident smirk on her lips as the light dances on her skin, creating a warm, inviting atmosphere.

Trainig details:

Trained only on videos.

• LR: 1e-5
• Optimizer: adamw_optimi
• epochs: 85
• dataset: 35 videos
• rank: 128
• batch size: 1
• gradient accumulation steps: 1

For training I used the Diffusion-Pipe repo.

Inference

It seems, the model likes long descriptive prompts. Look at the attached videos for prompt examples. Based on my tests if you use short prompts, the lora effect is weak.

For inference I recommend:

• ComfyUI native nodes
• Kijai's ComfyUI wrapper nodes

Trigger words: csetiarcane, Nfj1nx, blue hair

(If it isn't enough so you don't get the style/character, I recommend adding "animation style" to you prompt. It can help providing the style but in some cases the result will be too cartoony)

Strength: 0.9-1.0

Acknowledgement

Special thanks to Kijai for the tireless and excelent work done for the community to enable us to use these solutions as soon as possible, as well, as to Comfyanonymous for implementing them so quickly with such a good quality natively into ComfyUI, and also to TDRussel for making LoRA training available to the community so quickly.