High-fidelity character LoRA for Shiro-sama (シロ様) from The Helpful Fox Senko-san, trained on FLUX.2 Klein 9B 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, img2img, inpainting, and character replacements, with an emphasis on character likeness, clean linework, outfit flexibility, and descriptive natural-language prompting.<h2 id="trigger-word">Trigger word</h2><pre><code>shiro_sama</code></pre>Include <code>shiro_sama</code> in your positive prompt, followed by a description of the outfit, activity, setting, and visual style you want.<h2 id="what-this-model-does-best">What this model does best</h2><ul><li>Recognizable identity: Shiro’s reddish-pink eyes, long white hair, fox ears, short fluffy white tail, pink cheek markings, and catlike smile.</li><li>Outfit flexibility: Her canonical white-dress ensemble, jackets, casual clothing, seasonal outfits, and other wardrobe choices.</li><li>Descriptive prompting: Detailed control over clothing, expressions, poses, framing, lighting, and environments.</li><li>Image editing: Direct instructions for character changes or targeted corrections to an existing image.</li></ul><h2 id="recommended-generation-settings">Recommended generation settings</h2><ul><li>Base model: FLUX.2 Klein 9B Base</li><li>Trigger word: <code>shiro_sama</code></li><li>LoRA strength: Start at <code>1.0</code>; try <code>0.7–1.0</code> to balance likeness and flexibility</li><li>Sampler: Euler (<code>euler</code>)</li><li>Scheduler: Simple (<code>simple</code>)</li><li>CFG scale: Start at <code>7</code> for txt2img and img2img, then adjust for the task</li><li>Steps — faster generation: <code>28</code></li><li>Steps — balanced: <code>40</code></li><li>Steps — higher-detail option: <code>50</code></li><li>Resolution: Start at <code>1024 × 1024</code></li><li>Negative prompt: Completely empty</li></ul>Do not use a negative prompt with this LoRA. Leave the negative-prompt field completely empty.The examples below use 35 steps, with CFG 4.5 for txt2img and CFG 10 for targeted img2img edits. These are the actual example settings, separate from the general starting recommendation above.<h2 id="prompting-guidance">Prompting guidance</h2>Use direct, specific instructions. Start with <code>shiro_sama</code>, then describe the image in clear sentences.For txt2img, specify her clothing, expression, pose, camera angle, setting, and lighting. Reinforce important character features when needed, such as her white hair, white fox ears, reddish-pink eyes, pink cheek markings, and short fluffy white tail.For img2img and inpainting, clearly describe what should change and what should remain unchanged. A focused instruction can be enough for a small correction; broader character replacements benefit from a fuller description of her appearance.When color accuracy matters, describe Shiro’s hair, ears, and tail as white, with lighting-dependent highlights and shadows. Her eyes should remain reddish-pink, and her cheek markings pink.Use <code>short tail</code> to reinforce the intended tail length. Avoid <code>detailed eyes</code> if it causes her black pupils to disappear.<h2 id="default-outfit-and-clothing-changes">Default outfit and clothing changes</h2>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 <code>red shawl</code> rather than <code>red scarf</code>. A miko, kimono, or hakama outfit is an alternate costume.If <code>shiro_sama</code> alone does not recover her default outfit, reinforce it with:<pre><code>white dress, strapless dress, side slit, red shawl, black gloves, elbow gloves, black thighhighs, black choker</code></pre>These are garment descriptions; <code>shiro_sama</code> 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 or cardigan, omit conflicting garment descriptions and specify the new clothing directly.<h2 id="generation-examples">Generation examples</h2>All four examples use FLUX.2 Klein 9B Base, a LoRA weight of 1.0, 35 steps, Euler (<code>euler</code>), Simple (<code>simple</code>), and 1024 × 1024 resolution.Examples 1 and 2 have empty negative prompts. I forgot to clear the negative prompt for Examples 3 and 4, and those results did not come out as nicely. They are included for reference, but do not copy that part of their settings: leave your negative prompt empty.<h3 id="example-1:-autumn-walk-txt2img">Example 1: Autumn walk — txt2img</h3>Prompt<pre><code>shiro_sama shiro_sama with her fluffy white tail, white fox ears, reddish-pink eyes, pink cheek markings, red diamond pendant, and catlike mouth, taking a cheerful autumn walk, looking directly at the viewer with bright, relaxed eyes and a small, confident smile. Frame her closely from above her ears to mid-thigh, with her face as the main focus. Her body angles along the sidewalk while her head turns toward the viewer, caught mid-step with one shoulder slightly forward. She wears a short moss-green cardigan ending exactly at her waistband, hanging open over a cream mock-neck top, and a high-waisted plum pleated skirt. Simple knit texture, softly rounded sleeves, and a small amber leaf brooch. She holds a takeaway coffee below her chest, clear of her face, while her free arm swings naturally. Her hair and skirt move gently with her stride. Behind her on image-right is a softly lit Halloween shop window bordered by small orange and purple lights, with a few ghost and bat stickers. Inside, pumpkins and a plush ghost sit on a purple carpeted display. An autumn tree beside the shop frames the upper-left with orange-gold leaves. Warm late-afternoon sunlight catches her hair and shoulder, with gentle light across her face. Keep the shop softly detailed and its warm interior dimmer than her face, preserving a bright outdoor golden-hour atmosphere. Anime, anime coloring.</code></pre>Workflow: txt2img. LoRA weight: 1.0. Steps: 35. Sampler: Euler. Scheduler: Simple. CFG scale: 4.5. Seed: <code>308583758</code>. Size: 1024 × 1024. Negative prompt: Empty.<h3 id="example-2:-targeted-hand-correction-img2img">Example 2: Targeted hand correction — img2img</h3>Prompt<pre><code>shiro_sama shiro_sama's held out hand on the right side of the image should only have 4 fingers and a thumb</code></pre>This example uses a short, focused editing instruction to request a correction to the hand on image-right.Workflow: img2img edit, with one input image. LoRA weight: 1.0. Steps: 35. Sampler: Euler. Scheduler: Simple. CFG scale: 10. Seed: <code>2076514806</code>. Size: 1024 × 1024. Negative prompt: Empty.<h3 id="example-3:-autumn-walk-txt2img-with-an-accidental-negative-prompt">Example 3: Autumn walk — txt2img with an accidental negative prompt</h3>This generation uses the same positive prompt as Example 1, with a different seed and the accidental negative prompt reproduced below.Workflow: txt2img. LoRA weight: 1.0. Steps: 35. Sampler: Euler. Scheduler: Simple. CFG scale: 4.5. Seed: <code>326839520</code>. Size: 1024 × 1024.Negative prompt: The shared accidental negative prompt below. Leave this empty when generating your own images.<h3 id="example-4:-targeted-hand-correction-img2img-with-an-accidental-negative-prompt">Example 4: Targeted hand correction — img2img with an accidental negative prompt</h3>Prompt<pre><code>shiro_sama shiro_sama's held out hand on the right side of the image should only have 5 fingers.</code></pre>Workflow: img2img edit, with one input image. LoRA weight: 1.0. Steps: 35. Sampler: Euler. Scheduler: Simple. CFG scale: 10. Seed: <code>1063669291</code>. Size: 1024 × 1024.Negative prompt: The shared accidental negative prompt below. Leave this empty when generating your own images.<h3 id="accidental-negative-prompt-used-in-examples-3-and-4">Accidental negative prompt used in Examples 3 and 4</h3>The following is preserved only to document the settings used for those two examples. It is not a recommended negative prompt.<pre><code>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</code></pre><h2 id="dataset-and-caption-curation">Dataset and caption curation</h2>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.

Shiro-sama_NAI-V4.5-Curated (シロ様)

NovelAI 4.5 Curated Dataset - 598 Images

Shiro-sama_NAI-4.5-Curated-CAPTIONS.zip

Shiro-sama_NAI-V4_5-Curated_______flux2klein_9b_3397631_epoch_10.safetensors

High-fidelity character LoRA for Shiro-sama (シロ様) from The Helpful Fox Senko-san, for use with Krea 2 Turbo.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.<h2 id="trigger-word">Trigger word</h2><pre><code>shiro_sama</code></pre>Include <code>shiro_sama</code> in your positive prompt, followed by a description of the outfit, activity, setting, and visual style you want.<h2 id="what-this-model-does-best">What this model does best</h2><ul><li>Recognizable identity: Shiro’s reddish-pink eyes, long white hair, fox ears, short fluffy fox tail, and pink cheek markings.</li><li>Outfit flexibility: Her canonical white-dress ensemble, jackets, casual clothing, beach outfits, and other wardrobe choices.</li><li>Natural-language prompting: Supports descriptive prompts through Krea 2 Turbo’s Qwen3-VL-4B text encoder. Describe clothing, poses, expressions, lighting, and composition in clear sentences.</li><li>Scene flexibility: Indoor and outdoor scenes, traditional and modern settings, and different times of day.</li><li>Fast generation: Uses Krea 2 Turbo’s 8-step generation workflow for quick iteration.</li></ul><h2 id="recommended-generation-settings-comfyui">Recommended generation settings — ComfyUI</h2>Use an updated ComfyUI installation with native Krea 2 support.<ul><li>Base model: Krea 2 Turbo</li><li>Text encoder: Qwen3-VL-4B</li><li>Text encoder loader type: <code>krea2</code></li><li>VAE: Qwen Image VAE</li><li>Trigger word: <code>shiro_sama</code></li><li>Prompt style: Descriptive natural language</li><li>LoRA weight: Start at <code>1.0</code>; adjust within <code>0.85–1.0</code></li><li>Sampler: Euler (<code>euler</code>)</li><li>Scheduler: Simple (<code>simple</code>)</li><li>Steps: <code>8</code></li><li>CFG scale: <code>1.0</code> in ComfyUI</li><li>Denoise: <code>1.0</code> for <code>txt2img</code></li><li>Resolution: Start at <code>1024 × 1024</code>; Turbo also supports generation up to <code>2048 × 2048</code></li><li>Negative prompt: Leave empty at the recommended CFG</li></ul>Guidance setting: ComfyUI’s <code>CFG 1.0</code> disables additional classifier-free guidance. In Krea’s official inference code or Diffusers, use <code>CFG 0.0</code> / <code>guidance_scale=0.0</code> for the equivalent setting.For <code>img2img</code>, adjust denoise according to how much of the source image you want to preserve.<h2 id="adjusting-lora-strength">Adjusting LoRA strength</h2>Start at <code>1.0</code> and adjust in small increments.<ul><li>Try <code>0.85–1.0</code> to balance character identity with outfit and style flexibility.</li><li>Lower the weight if character details overpower the requested clothing or composition.</li><li>Try <code>1.1–1.2</code> for stronger character influence, checking for exaggerated features or reduced flexibility.</li></ul><h2 id="prompting">Prompting</h2>Use descriptive natural language. Start with <code>shiro_sama</code>, then explain the image in clear sentences.<ul><li>Describe her clothing and expression explicitly.</li><li>Specify her pose or activity, including what her hands are doing.</li><li>Include the setting, camera angle, framing, and lighting.</li><li>State the intended style, such as anime illustration, clean linework, or soft cel shading.</li><li>Keep instructions consistent and prioritize the details that matter most.</li></ul>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.<h2 id="default-outfit-and-clothing-changes">Default outfit and clothing changes</h2>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 <code>red shawl</code> rather than <code>red scarf</code>. A miko, kimono, or hakama outfit is an alternate costume.If <code>shiro_sama</code> alone does not recover her default outfit, reinforce it with:<pre><code>white dress, strapless dress, side slit, red shawl, black gloves, elbow gloves, black thighhighs, black choker</code></pre>These are garment descriptions; <code>shiro_sama</code> 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 <code>strapless dress</code> and describe the jacket directly. To keep the underlying dress and black accessories, retain <code>white dress, side slit, black thighhighs, black choker</code>, then add the jacket description.Use <code>short tail</code> to reinforce the intended tail length. Avoid <code>detailed eyes</code> if it causes her black pupils to disappear.<h2 id="generation-examples">Generation examples</h2>The following examples were generated with Krea 2 Turbo v1.0, using 15 steps rather than the recommended 8-step starting configuration.Shared settings: txt2img, LoRA weight 1.0, Euler (<code>euler</code>), CFG 1.0, and 1024 × 1024 resolution.<h3 id="example-1:-sunglasses-above-her-forehead">Example 1: Sunglasses above her forehead</h3>Prompt<pre><code>anime, shiro_sama on a sunny airport tarmac near Miami, caught mid-step as she begins pulling her suitcase toward a small passenger airplane. Full-body view at chest height, fluffy white fox tail just above her luggage, her body still facing the viewer, one foot stepping diagonally toward the boarding area and the opposite heel lifting, just before she turns. Her raised right hand holds slightly squared plum sunglasses by their outer corner just in front of her forehead, caught as she begins lowering them into place. The entire frame and both lenses remain above her eyebrows, leaving her eyes completely unobstructed. Her face points toward the viewer, both eyes making direct eye contact, with slightly lowered upper lids, relaxed eyebrows, a subtly raised chin, and a confident, playful smile. Her left hand pulls the extended suitcase handle, the case tilted onto its rear wheels and just beginning to roll behind her. She wears an ivory short-sleeved camp-collar blouse with narrow plum piping, loosely tucked into high-waisted sea-glass-teal ankle-length trousers, a slim tan belt, and ivory low-top sneakers with tan soles. Her hair tips and blouse hem lift slightly with her movement and a light breeze, her tail following behind without obscuring the luggage. The compact warm-yellow suitcase faces the viewer, displaying a cute curled-up fox sticker, a smaller blue wave sticker, and a coral travel stamp. Behind her stands a well-kept pearl-white twin-engine propeller passenger plane with a navy-and-teal stripe, oval cabin windows, stationary propellers, and short boarding stairs. A few distant passengers approach and ascend the steps with their carry-ons. Palm trees stand off to the left beside a low terminal, beneath a vivid blue sky with scattered white clouds. Upper-left sunlight peeks through palm fronds in a small sunburst, casting warm highlights across her sunglasses, shoulder, and suitcase, with soft blue-violet shadows on the pale tarmac. Keep her eyes and smile clearly visible, the background gently softened, and the sunny atmosphere warm and breezy without harsh glare.</code></pre>Workflow: txt2img. Checkpoint: Krea 2 Turbo v1.0. LoRA weight: 1.0. Steps: 15. Sampler: Euler (<code>euler</code>). CFG scale: 1.0. Seed: <code>813028192</code>. Size: 1024 × 1024.<h3 id="example-2:-looking-above-the-sunglasses">Example 2: Looking above the sunglasses</h3>Prompt<pre><code>anime, shiro_sama on a sunny airport tarmac near Miami, caught mid-step as she begins pulling her suitcase toward a small passenger airplane. Full-body view at chest height, fluffy white fox tail just above her luggage, her body still facing the viewer, one foot stepping diagonally toward the boarding area and the opposite heel lifting, just before she turns. Her right hand guides slightly squared plum sunglasses onto her nose, holding the outer frame corner with the lenses still below her eyes, looking directly at the viewer above the rims. Her chin is slightly raised, eyebrows relaxed, and mouth curved into a confident, slightly asymmetric smile. Her left hand pulls the extended suitcase handle, the case tilted onto its rear wheels and just beginning to roll behind her. She wears an ivory short-sleeved camp-collar blouse with narrow plum piping, loosely tucked into high-waisted sea-glass-teal ankle-length trousers, a slim tan belt, and ivory low-top sneakers with tan soles. Her hair tips and blouse hem lift slightly with her movement and a light breeze, her tail following behind without obscuring the luggage. The compact warm-yellow suitcase faces the viewer, displaying a cute curled-up fox sticker, a smaller blue wave sticker, and a coral travel stamp. Behind her stands a well-kept pearl-white twin-engine propeller passenger plane with a navy-and-teal stripe, oval cabin windows, stationary propellers, and short boarding stairs. A few distant passengers approach and ascend the steps with their carry-ons. Palm trees stand off to the left beside a low terminal, beneath a vivid blue sky with scattered white clouds. Upper-left sunlight peeks through palm fronds in a small sunburst, casting warm highlights across her sunglasses, shoulder, and suitcase, with soft blue-violet shadows on the pale tarmac. Keep her eyes and smile clearly visible, the background gently softened, and the sunny atmosphere warm and breezy without harsh glare.</code></pre>Workflow: txt2img. Checkpoint: Krea 2 Turbo v1.0. LoRA weight: 1.0. Steps: 15. Sampler: Euler (<code>euler</code>). CFG scale: 1.0. Seed: <code>915615285</code>. Size: 1024 × 1024.<h3 id="shared-negative-prompt">Shared negative prompt</h3>Both examples contain the following negative prompt in their Civitai generation metadata:<pre><code>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</code></pre>For the recommended ComfyUI CFG 1.0 workflow, leave the negative prompt empty because additional classifier-free guidance is disabled. The text above is retained as a record of the examples’ submitted generation settings.<h2 id="dataset-and-caption-curation">Dataset and caption curation</h2>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.

Shiro-sama_NAI-V4_5-Curated_______krea2_3397676_epoch_10.safetensors

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.<h2 id="trigger-word">Trigger word</h2><pre><code>shiro_sama
</code></pre>Include <code>shiro_sama</code> 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.<h2 id="what-this-model-does-best">What this model does best</h2><ul><li>Recognizable identity: Shiro’s reddish-pink eyes, long white hair, fox ears, short fluffy fox tail, and pink cheek markings. </li><li>Outfit flexibility: Her canonical white-dress ensemble, jackets, casual clothing, beach outfits, and other wardrobe choices. </li><li>Expressive scenes: Varied expressions, poses, activities, and compositions. </li><li>Flexible environments: Indoor and outdoor scenes, traditional and modern settings, and different times of day. </li><li>Descriptive prompting: Natural-language control over clothing, expression, pose, camera position, lighting, and environment. </li></ul><h2 id="prompting">Prompting</h2>Use descriptive natural language. Start with the trigger word, then explain the image in clear sentences.<ul><li>Describe her clothing and expression explicitly. </li><li>Specify her pose or activity, including what her hands are doing. </li><li>Include the setting, camera angle, framing, and lighting. </li><li>State the intended style, such as anime illustration, clean linework, or soft cel shading. </li><li>Keep instructions consistent and prioritize the details that matter most. </li></ul>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.<h2 id="default-outfit-and-clothing-changes">Default outfit and clothing changes</h2>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 <code>red shawl</code> rather than <code>red scarf</code>. A miko, kimono, or hakama outfit is an alternate costume.If <code>shiro_sama</code> alone does not recover her default outfit, reinforce it with:<pre><code>white dress, strapless dress, side slit, red shawl, black gloves, elbow gloves, black thighhighs, black choker
</code></pre>These are garment descriptions; <code>**shiro_sama**</code> 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 <code>strapless dress</code> and describe the jacket directly.Use <code>short tail</code> to reinforce the intended tail length. Avoid <code>detailed eyes</code> if it causes her black pupils to disappear.<h2 id="model-setup-comfyui">Model setup — ComfyUI</h2>Use an updated ComfyUI installation with native Boogu support.<ul><li>Base model: <code>Boogu-Image-0.1-Base</code> </li><li>Text encoder: Qwen3-VL-8B, such as <code>qwen3vl_8b_fp8_scaled.safetensors</code> </li><li>Text encoder loader type: <code>boogu</code> </li><li>VAE: FLUX.1 VAE, such as <code>flux1_vae_bf16.safetensors</code> </li><li>Trigger word: <code>shiro_sama</code> </li></ul><h2 id="suggested-starting-settings">Suggested starting settings</h2><ul><li>LoRA weight: Start at <code>1.0</code> </li><li>Sampler: Euler (<code>euler</code>) </li><li>Scheduler: Simple (<code>simple</code>) </li><li>Steps: Start at <code>35</code>; explore <code>25–50</code> </li><li>CFG scale: Start at <code>3.5</code>; explore <code>2.0–5.0</code> </li><li>Resolution: Start at <code>1024 × 1024</code>; the examples below use <code>1216 × 832</code> </li><li>Denoise: <code>1.0</code> for <code>txt2img</code> </li><li>Negative prompt: Start empty; add targeted exclusions if needed </li></ul>For <code>img2img</code>, adjust denoise according to how much of the source image you want to preserve.<h2 id="adjusting-lora-strength">Adjusting LoRA strength</h2>Start at <code>1.0</code> and adjust in small increments.<ul><li>Try <code>0.85–1.0</code> to balance character identity with outfit and style flexibility. </li><li>Lower the weight if character details overpower the requested clothing or composition. </li><li>Try <code>1.1–1.2</code> for stronger character influence, checking for exaggerated features or reduced flexibility. </li></ul>All three examples below use a LoRA weight of 1.0.<h2 id="negative-prompting">Negative prompting</h2>For a minimal negative prompt, start with:<pre><code>worst quality, low quality, blurry, deformed, text, watermark
</code></pre>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.<h2 id="examples:-diving-for-a-beach-ball">Examples: Diving for a beach ball</h2>These three generations use the same positive and negative prompts, with different seeds and step counts.<h3 id="prompt">Prompt</h3><pre><code>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.
</code></pre><h3 id="negative-prompt">Negative prompt</h3><pre><code>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
</code></pre><h3 id="generation-settings">Generation settings</h3>All three examples use txt2img, Boogu Image, a LoRA weight of 1.0, Euler (<code>**euler**</code>), CFG 3.5, and 1216 × 832 resolution.<ul><li>Example 1: 50 steps. Seed: <code>1288911761</code>. </li><li>Example 2: 35 steps. Seed: <code>606058672</code>. </li><li>Example 3: 35 steps. Seed: <code>1325828615</code>.</li></ul><h2 id="dataset-and-caption-curation">Dataset and caption curation</h2>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.

Files with SHA256: d8372a76dfdd7f3b4ad741ad5ed9db3299367fe9103199b9384f8f2a10da0e5d