Sayuri LoRA (Blue Archive)

Training Information
- Dataset: 30 Images
- Epochs: 10
- Total Steps: 1800
- Batch Size: 1
- Repeats: 6

Recommended Model
- Epoch 8 (Recommended)

"Please use the V3 LoRA!"

Sayuri (サユリ) - Blue Archive

⚙️ Recommended Settings

• Weight: 0.8 ~ 1.0

• Trigger Words: sr

• Clip Skip: 2

• Sampling method: Euler a

• Steps: 28 ~ 30

• Width, Height: 1024,1024

• CFG Scale: 5 ~ 7

🔍 Recommended Hires. fix Settings

• Upscaler: R-ESRGAN 4x+ Anime6

• Hires steps: 10

• Denoising strength: 0.4

• Upscale by: 2

Recommended Settings

- ADetailer: ON

- Upscaling: 1.5x ~ 2x

💡 Prompt Example (Base Outfit)

(masterpiece:1.1), pastel colors,
detached collar, blue bowtie, white swimsuit, one-piece swimsuit, halterneck, bare shoulders, white armband, white wristwatch, thigh strap, thigh holster, bare legs, barefoot,
sr, 1girl, solo, white hair, long hair, bangs, hair between eyes, sidelocks, white eyelashes, sayuri_eyes, blue eyes, halo, x hair ornament, hair clip, orca hair ornament, 
<lora:Sayuri_v3:0.8>

If the shadows are too deep, try adding "backlighting, rim light" to the negative prompt.

🚫 Negative Prompt Example

(worst quality, low quality:1.2), (lowres:1.2), bad anatomy, (bad hands:1.2), (cast shadow, hard shadows:1.2), directional lighting, chiaroscurotext, error, missing fingers, extra digit, fewer digits, cropped, jpeg artifacts, signature, watermark, username