Description

This is a WAI / Illustrious-SDXL character outfit LoRA for Vendela - Rose Mystery / 刺玫「蔷薇之谜」, trained with the custom trigger vendela_rose and the character anchor vendela (arknights).

It is designed to recall:

• long peach-pink hair and amber eyes

• fluffy cat ears and fluffy brown cat tail

• silver wire halo and white rose hair ornament

• white lace princess dress with layered frilled skirt

• green vine straps around ankles and calves

• bare feet, white roses and rose vine staff

Training metadata records SFW samples: 100, NSFW samples: 0. The project configuration allowed NSFW generation, but the final selected training dataset for this run is SFW only.

Recommended Usage

Default Weight

0.70

Useful Range

0.65-0.85

Motion / Composition First

0.60 - 0.75

Avoid as Default

0.95 - 1.0

Basic Prompt

<lora:vendela_rose_recommended_e000008:0.70>, vendela_rose, vendela (arknights), 1girl, solo, no text, looking at viewer, upper body, soft smile, simple background

Full Character Prompt

<lora:vendela_rose_recommended_e000008:0.70>, masterpiece, best quality, amazing quality, very aesthetic, absurdres, vendela_rose, vendela (arknights), 1girl, solo, long peach pink hair, amber eyes, fluffy cat ears, fluffy brown cat tail, silver wire halo, white rose hair ornament, white lace princess dress, layered frilled white skirt, green vine straps around ankles and calves, bare feet, rose vine staff, white roses, no text, detailed outfit, soft lighting

Negative Prompt

low quality, worst quality, lowres, bad anatomy, bad hands, bad fingers, extra fingers, missing fingers, extra limbs, disfigured, deformed, blurry, jpeg artifacts, text, watermark, logo, signature, arknights logo, chinese text, english text, caption, artist name, bad face, bad eyes, multiple girls, 2girls, cropped head, out of frame, wrong hair color, wrong eye color, wrong outfit, missing cat ears, missing cat tail, missing silver halo, missing white rose hair ornament, missing vine straps, shoes, socks, boots, armor, armored knight, multiple characters

Example Prompts

Portrait

<lora:vendela_rose_recommended_e000008:0.70>, vendela_rose, vendela (arknights), 1girl, solo, no text, long peach pink hair, amber eyes, fluffy cat ears, fluffy brown cat tail, silver wire halo, white rose hair ornament, white lace princess dress, layered frilled white skirt, green vine straps around ankles and calves, bare feet, rose vine staff, white roses, portrait, close-up, looking at viewer, beautiful face, detailed amber eyes, simple rose background

Full Body

<lora:vendela_rose_recommended_e000008:0.70>, vendela_rose, vendela (arknights), 1girl, solo, no text, long peach pink hair, amber eyes, fluffy cat ears, fluffy brown cat tail, silver wire halo, white rose hair ornament, white lace princess dress, layered frilled white skirt, green vine straps around ankles and calves, bare feet, rose vine staff, white roses, full body, standing, elegant pose, clear outfit, clear bare feet, simple background

Seated / Sensual

<lora:vendela_rose_recommended_e000008:0.70>, vendela_rose, vendela (arknights), 1girl, solo, no text, long peach pink hair, amber eyes, fluffy cat ears, fluffy brown cat tail, silver wire halo, white rose hair ornament, white lace princess dress, layered frilled white skirt, green vine straps around ankles and calves, bare feet, rose vine staff, white roses, sitting, crossed legs, leaning back, charming smile, soft warm lighting, elegant lounge

Recommended Generation Settings

Checkpoint: WAI-Nsfw-Illustrious-17
Resolution: 1024 x 1024
Sampler: DPM++ 2M
Scheduler: Karras
Steps: 30
CFG scale: 6.5
LoRA weight: 0.70

Known Limitations

• The LoRA was trained and evaluated for WAI / Illustrious-SDXL. Other model families may change identity and outfit details.

• Very high weights may strengthen the dress and rose motif but can reduce pose flexibility.

• Complex full-body poses may need lower weight and explicit tags for cat ears, cat tail, bare feet and green vine straps.

• Strong outfit LoRAs may conflict with the white lace dress and vine strap design.

Training Summary

Training images: 100 PNG
Caption files: 100 TXT
SFW samples: 100
NSFW samples: 0
Base model: WAI-Nsfw-Illustrious-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
Recommended file: vendela_rose_recommended_e000008.safetensors
Recommended weight: 0.70