<h1 id="senko-san-illustrious-xl-v0.1-(hd-ultra-version)">Senko-san — Illustrious XL v0.1 (HD ULTRA Version)</h1>High-fidelity character LoRA for Senko-san from The Helpful Fox Senko-san, trained on Illustrious XL v0.1.This LoRA was trained on a high-quality synthetic dataset of 1,000 crisp, native 1024×1024 HD images generated with NovelAI 5.0 Curated. It produces images with a sharp, highly detailed look with precise character features and high-resolution rendering. It is designed for standalone character generation (<code>txt2img</code>) as well as <code>img2img</code>, inpainting, and character face/head replacement workflows.This is the ULTRA version because it was trained on a dataset that includes both the original 598-image base dataset and an all-new 402-image activities dataset.(Note: If you want the softer, standard-definition anime screencap look, check out my separate SD ANIME Senko-san LoRA (coming soon). This release is built specifically for modern, maximally sharp, high-resolution generation).<hr /><h2 id="quick-generation-settings">Quick Generation Settings</h2>Setting Recommended Value Base Model Illustrious XL v0.1 (or compatible checkpoints) Trigger Tag <code>senko_san</code> Helpful Tags <code>1girl, solo, hair ornament</code> LoRA Weight <code>0.85 – 1.0</code> (<code>0.85</code> for outfit flexibility, <code>1.0</code> for full likeness) Sampler <code>Euler a</code> Steps <code>28</code> CFG Scale <code>7</code> Resolution <code>1024 × 1024</code> Clip Skip <code>2</code><h3 id="lora-weight">LoRA Weight</h3><ul><li><code>0.85</code> — good character likeness with more room for outfit and scene control.</li><li><code>0.90</code> — balanced likeness and prompt flexibility.</li><li><code>1.0</code> — strongest Senko likeness.</li><li><code>0.70</code> — still usable, with a softer and less forceful character likeness.</li></ul><h3 id="negative-prompt">Negative Prompt</h3>Heavy negative prompting is not required. A simple negative prompt is sufficient:<pre><code>worst quality, low quality, blurry
</code></pre>Optional extended negative if you encounter artifacts:<pre><code>lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, artifacts, signature, watermark, username, blurry
</code></pre><hr /><h2 id="example-prompts">Example Prompts</h2><h3 id="1.-traditional-shrine-maiden-kimono-(txt2img)">1. Traditional Shrine Maiden / Kimono (<code>txt2img</code>)</h3><blockquote>Prompt: <code>masterpiece, best quality, senko_san, 1girl, fox ears, kitsune, kimono, smiling, tea cup, traditional japanese room</code> Weight: <code>1.0</code> | Steps: <code>28</code> | Sampler: <code>Euler a</code> | CFG: <code>7</code></blockquote><h3 id="2.-modern-streetwear-(txt2img)">2. Modern Streetwear (<code>txt2img</code>)</h3><blockquote>Prompt: <code>masterpiece, best quality, senko_san, 1girl, fox ears, kitsune, black leather motorcycle jacket, blue jeans, city street, night, neon lights</code> Weight: <code>0.85</code> | Steps: <code>28</code> | Sampler: <code>Euler a</code> | CFG: <code>7</code></blockquote><h3 id="3.-casual-morning-(txt2img)">3. Casual Morning (<code>txt2img</code>)</h3><blockquote>Prompt: <code>masterpiece, best quality, senko_san, 1girl, fox ears, oversized white hoodie, holding coffee mug, modern kitchen, morning light</code> Weight: <code>0.90</code> | Steps: <code>28</code> | Sampler: <code>Euler a</code> | CFG: <code>7</code></blockquote><h3 id="4.-img2img-inpainting-replacement">4. Img2img / Inpainting Replacement</h3><blockquote>Prompt: <code>senko_san, 1girl, looking at viewer, smile, hair ornament</code> Weight: <code>0.85–1.0</code> | Denoise: <code>0.55–0.70</code> | Resolution: <code>1024 × 1024</code></blockquote><hr /><h3 id="dataset-and-tagging-procedure">Dataset &amp; Tagging Procedure</h3>The dataset contains 1,000 high-resolution 1024×1024 images.Tagging followed a multi-stage workflow combining two automated taggers, human review, custom tooling, and AI-assisted policy iteration:<ol><li>Two-Stage Auto-Tagging:<ul><li>Step 1: An initial auto-tagging pass was performed using Civitai's built-in auto-tagger.</li><li>Step 2: A subsequent pass was run using the new PixAI auto-tagger (available on Hugging Face).</li></ul></li><li>Manual Review &amp; Baseline Policy:<ul><li>A representative set of 102 images was manually reviewed by hand to establish the initial tagging dictionary and baseline rules.</li><li>An initial draft policy was created interactively with an AI agent based on these reviewed images.</li></ul></li><li>Custom Tooling:<ul><li>Tag curation was managed using a customized version of toshiaki1729's dataset editor, modified to allow AI agents to assist with dataset review and tag management.</li><li>The tool takes the raw auto-tagger output, applies the active policy rules, and outputs the curated tag set.</li></ul></li><li>AI-Assisted Policy Iteration:<ul><li>The auto-tagger outputs were loaded into the custom editor.</li><li>An AI agent analyzed the output to surface edge cases, normalize inconsistent terminology, and categorize tags under explicit policy rules: Keep, Reject, or Unknown.</li><li>As new cases were surfaced, the policy was refined interactively and reapplied across the dataset to strip noise, prevent concept bleed, and produce the final curated tag set.</li></ul></li></ol><hr /><h2 id="training-specifications">Training Specifications</h2>| Parameter | Setting ||---|---|| Training Engine | <code>ostris/ai-toolkit</code> || Base Model | Illustrious XL v0.1 || Architecture | SDXL || Precision | BF16 || Training Steps | 2,000 || Batch Size | 2 || Gradient Accumulation | 1 || Network Rank / Alpha | Dim 32 / Alpha 32 || Optimizer | Automagic || UNet Learning Rate | <code>1e-4</code> || Text Encoder Learning Rate | <code>1e-5</code> || LR Scheduler | Cosine || Tag Shuffling | Enabled || Preserve Tag | 1 | (Preserve Tag 1: <code>senko_san</code>)| Tag Dropout | 0.05 || Clip Skip | 2 |

Senko-san_NAI-V5-Curated_HD_ULTRA

NovelAI 5.0 Curated Dataset - 1,000 HD Images

Files with SHA256: 39429a19cb2c910c71d94aa485d5bdf331d963f297e73a4ad1eef1434605f4c7