Trigger Words

4zum4y4ch1h0, 1girl, long hair, twintails, blue hair, yellow eyes, medium breasts, hair ribbon, white ribbon,

Outfits

military uniform, black uniform, gloves, white gloves, belt, long skirt, black skirt, frills, boots, black footwear

📸 Preview Prompts

anime coloring, masterpiece, best quality, amazing quality, 4zum4y4ch1h0, 1girl, long hair, twintails, blue hair, yellow eyes, medium breasts, hair ribbon, white ribbon, military uniform, black uniform, gloves, white gloves, belt, long skirt, black skirt, frills, boots, black footwear, solo, pose, standing, looking at viewer, from front, ((((white background, simple background))))

------------------------------------------------------------------------------------------

⚙️ Core Settings

Steps: 25

Use 20–24 for faster drafts, and 26–30 for final renders when you want more polish and definition.

Sampler: Euler a

A stable and fast baseline sampler, ideal for detailed prompts and character-focused compositions.

Schedule Type: Automatic

Lets the interface choose the best scheduler behavior for your selected sampler.

CFG Scale: 5

Lower values (4–5) give more freedom and creative variation.

Higher values (6–7) enforce the prompt more strictly, but may cause oversaturation or stiffness.

🧩 Model & Precision

Model/Checkpoint: WAI-NSFW-illustrious-SDXL (v16).

Clip Skip: 2

Common in anime and merged-style models. It emphasizes the main subject and keeps outlines clean. If you are getting bad results with Forge try to set it at 1.

ENSD: 31337

Ensures consistent behavior across multiple sessions and tests.

Token Merging Ratio: 0.5

Speeds up generation by merging less-important tokens. If you notice detail loss or visual drift, reduce it to 0.25 or disable it completely.

VAE: Not needed, already embedded in the model/checkpoint

Specifies the active VAE (Variational Autoencoder) used for color space and tonal compression.

👤 ADetailer - Face (2nd Pass)

Detector Model: face_yolov8n.pt
Confidence: 0.3
Dilate/Erode: 4
Mask Blur: 4
Denoising Strength: 0.4
Inpaint Only Masked: True
Inpaint Padding: 32
Purpose: Provides gentle, localized facial refinement, improving facial definition without altering lighting, color balance, or the overall composition.

⚠️ Important Note (Very Common Mistake): If you may notice that in your result image the character has the wrong eye color. This happens because ADetailer is applying the face enhancement on the 1st pass, when it should actually be applied on the 2nd pass.

When ADetailer enhances the face too early (1st pass), it can unintentionally modify essential character features such as:

eye color
facial expression
face shape
other identity-defining traits

To avoid this, it’s best to apply the face enhancement on the 2nd pass, ensuring the model has already established the correct characteristics before refinement.

🆙 HiRes Settings

HiRes Upscale: 1.5

Upscales the image to 150% of its original resolution. Use 1.5x for a moderate resolution boost that enhances details without excessive VRAM usage.

HiRes Steps: 20

Defines the number of steps during upscaling. 20 steps offer a good balance between quality and processing time, improving detail while avoiding excessive rendering time.

HiRes Upscaler: R-ESRGAN 4x+ Anime6B

An upscaler designed for anime-style images, perfect for enhancing details while keeping clean lines. Ideal for illustrations, not recommended for photo-realistic images.

Denoising Strength: 0.35-0.5

Controls noise removal during upscaling. Set between 0.35–0.5 to remove noise while preserving details, use 0.4 for a balanced result.

⚙️ Generation Settings

Azuma Yachiho [Mato Seihei no Slave] - v1.0