Highmore 及山海 outfit LoRA for WAI v17 / Illustrious-SDXL. Use highmore_ji_shanhai with highmore (arknights) around weight 0.80; recommended range is 0.7-0.9.

Source / Character / Skin
Highmore / 海沫 from Arknights, skin 及山海.

Trigger Words
highmore_ji_shanhai
highmore (arknights)
1girl

Recommended Usage
<lora:highmore_ji_shanhai_strict_nai_recommended_e000008:0.80>, highmore_ji_shanhai, highmore (arknights), 1girl, solo, no text, looking at viewer

Character / Full Outfit Prompt
<lora:highmore_ji_shanhai_strict_nai_recommended_e000008:0.80>, masterpiece, best quality, amazing quality, very aesthetic, highmore_ji_shanhai, highmore (arknights), 1girl, solo, long wavy hair, very long dark purple hair, black purple gradient hair, flowing hair, side-swept bangs, hair between eyes, long sidelocks, blue fin-like hair ornament, purple hairclip, branch antlers, pointed ears, light blue eyes, pearl earrings, high collar, pendant, dark navy strapless shoulder, floral embroidery, white short skirt, asymmetrical dress, long translucent white shawl, blue purple long cape, tassel ornament, bare shoulders, bare legs, no socks, slender legs, pointed toe high heels, gray pumps, black ankle strap shoes, gold buckle, blue ribbon on shoes, elegant heels, solo, no text

Condensed Negative Prompt
low quality, bad , blurry, bad proportions, deformed hands, extra limbs, extra fingers, disfigured, low resolution, oversaturated, grainy, poorly drawn face, text, watermark, low detail, bad lighting, noisy background, bad composition, wrong perspective, long neck, asymmetrical, pixelated, distorted, out of focus, artifacts, different outfit, alternate outfit, outfit change, wrong outfit, incorrect outfit, mismatched outfit, incomplete outfit, censored, bad face, bad eyes, cropped face, cropped head, multiple girls, duplicate

Recommended Generation Settings
Checkpoint: -17
Resolution: 1024 x 1024
Sampler: DPM++ 2M
Scheduler: Karras
Steps: 30
CFG scale: 6.5
LoRA weight: 0.80

Training Summary
Training images: 100 PNG
Caption files: 100 TXT
SFW samples: 100
samples: 0
Base model: -17
Training type: SDXL LoRA, U-Net only
Network dim / alpha: 32 / 16
Network dropout: 0.05
Optimizer: AdamW8bit
Precision: bf16
Recommended checkpoint: epoch 8