Shiro-sama_NAI-V4.5-Curated (シロ様)

High-fidelity character LoRA for Shiro-sama (シロ様) from The Helpful Fox Senko-san, for use with Boogu-Image-0.1 Base.

This LoRA was built from a high-quality synthetic dataset of 598 captioned images generated with NovelAI 4.5 Curated. It is designed for txt2img character generation and img2img, with an emphasis on character likeness, clean linework, outfit flexibility, and descriptive natural-language prompting.

Trigger word

shiro_sama

Include shiro_sama in your positive prompt, followed by a description of the outfit, activity, setting, and visual style you want. Use the Base model. Turbo is not recommended for this LoRA.

What this model does best

• Recognizable identity: Shiro’s reddish-pink eyes, long white hair, fox ears, short fluffy fox tail, and pink cheek markings.
  

• Outfit flexibility: Her canonical white-dress ensemble, jackets, casual clothing, beach outfits, and other wardrobe choices.
  

• Expressive scenes: Varied expressions, poses, activities, and compositions.
  

• Flexible environments: Indoor and outdoor scenes, traditional and modern settings, and different times of day.
  

• Descriptive prompting: Natural-language control over clothing, expression, pose, camera position, lighting, and environment.
  

Prompting

Use descriptive natural language. Start with the trigger word, then explain the image in clear sentences.

• Describe her clothing and expression explicitly.
  

• Specify her pose or activity, including what her hands are doing.
  

• Include the setting, camera angle, framing, and lighting.
  

• State the intended style, such as anime illustration, clean linework, or soft cel shading.
  

• Keep instructions consistent and prioritize the details that matter most.
  

For dynamic scenes, describe the direction of movement and how her body, hair, clothing, and tail respond. For precise compositions, explain her position relative to the camera and any foreground objects.

Default outfit and clothing changes

Shiro’s default outfit is her white strapless dress and red shawl ensemble: a white dress with side slits and a gold crescent motif on the bodice, a deep-red shawl draped around her upper arms, black elbow-length gloves, black thigh-high stockings, and a broad black choker with a red diamond-shaped ornament.

For the default wrap, use red shawl rather than red scarf. A miko, kimono, or hakama outfit is an alternate costume.

If shiro_sama alone does not recover her default outfit, reinforce it with:

white dress, strapless dress, side slit, red shawl, black gloves, elbow gloves, black thighhighs, black choker

These are garment descriptions; **shiro_sama** remains the character trigger.

For alternate outfits, explicitly describe the replacement clothing. Retain descriptions only for the default pieces you want to keep. When adding a jacket, omit strapless dress and describe the jacket directly.

Use short tail to reinforce the intended tail length. Avoid detailed eyes if it causes her black pupils to disappear.

Model setup — ComfyUI

Use an updated ComfyUI installation with native Boogu support.

• Base model: Boogu-Image-0.1-Base
  

• Text encoder: Qwen3-VL-8B, such as qwen3vl_8b_fp8_scaled.safetensors
  

• Text encoder loader type: boogu
  

• VAE: FLUX.1 VAE, such as flux1_vae_bf16.safetensors
  

• Trigger word: shiro_sama
  

Suggested starting settings

• LoRA weight: Start at 1.0
  

• Sampler: Euler (euler)
  

• Scheduler: Simple (simple)
  

• Steps: Start at 35; explore 25–50
  

• CFG scale: Start at 3.5; explore 2.0–5.0
  

• Resolution: Start at 1024 × 1024; the examples below use 1216 × 832
  

• Denoise: 1.0 for txt2img
  

• Negative prompt: Start empty; add targeted exclusions if needed
  

For img2img, adjust denoise according to how much of the source image you want to preserve.

Adjusting LoRA strength

Start at 1.0 and adjust in small increments.

• Try 0.85–1.0 to balance character identity with outfit and style flexibility.
  

• Lower the weight if character details overpower the requested clothing or composition.
  

• Try 1.1–1.2 for stronger character influence, checking for exaggerated features or reduced flexibility.
  

All three examples below use a LoRA weight of 1.0.

Negative prompting

For a minimal negative prompt, start with:

worst quality, low quality, blurry, deformed, text, watermark

Add targeted negatives when needed for ears, tails, anatomy, clothing, or other recurring problems. The beach examples below use a longer negative prompt, reproduced in full.

Examples: Diving for a beach ball

These three generations use the same positive and negative prompts, with different seeds and step counts.

Prompt

shiro_sama Create a dynamic anime beach scene featuring the shiro_sama wearing a loose, waist-length white beach blouse with short flutter sleeves, a high overlapping neckline edged in plum purple, and a small red ribbon fastening it at one side. A tiny purple fox-paw motif decorates the lower hem. Pair it with her red wraparound beach skirt, purple sunglasses above her forehead, and orange wrist scrunchies. Capture her diving forward to knock a blue-and-white beach ball back into play, her body stretched almost horizontally above the sand. The sleeves and hem of her blouse lift slightly as she moves. One arm reaches toward the ball with a closed fist while the other extends behind her for balance. Her long white hair and fluffy tail stream backward with the movement, and her upright ears, reddish-pink eyes, pink cheek markings, and excited open-mouthed smile remain clearly visible. Use a low camera angle close to the sand, with her reaching hand enlarged by perspective and the beach ball partly cropped in the upper-right foreground. A sloping sandy beach fills the lower frame, with dense green coastal trees beneath a vivid blue sky and scattered white clouds. Use bright colors, clean anime linework, and crisp summer lighting.

Negative prompt

tears, crying, cry, sweat, rain, worst quality, low quality, lowres, jpeg artifacts, bad anatomy, bad hands, extra fingers, missing fingers, fused fingers, extra arms, extra legs, disconnected limbs, multiple tails, detached tail, cat ears, human ears, photorealistic, 3d, cgi, realistic skin, painterly, gradient shading, monochrome, muted colors, underexposed, blurry face, multiple girls, crowd, child, loli, petite, text, watermark, signature, long tail, belt

Generation settings

All three examples use txt2img, Boogu Image, a LoRA weight of 1.0, Euler (**euler**), CFG 3.5, and 1216 × 832 resolution.

• Example 1: 50 steps. Seed: 1288911761.
  

• Example 2: 35 steps. Seed: 606058672.
  

• Example 3: 35 steps. Seed: 1325828615.

Dataset and caption curation

The dataset contains 598 tagged synthetic images generated with NovelAI 4.5 Curated. Caption curation combined PixAI auto-tagging with manual review, custom tooling, and AI-assisted policy refinement. The tagging dictionary and baseline rules were carried over from my previous Senko-san LoRA projects, then revised and refined for Shiro.

Tag curation used a customized version of toshiaki1729’s dataset editor, modified to support AI-assisted dataset review and tag management. The editor applied the active policy to the raw auto-tagger output and produced curated captions.

AI-assisted review surfaced edge cases, normalized inconsistent terminology, and classified tags as Keep, Reject, or Unknown. The policy was refined and reapplied across the dataset to reduce caption noise and concept bleed.