Trained on high-quality CGs from the adult visual novel Nie no Hakoniwa.All showcase images were generated text-to-image (t2i) using this LoRA on the official Anima base v1.0.<ul><li>Resolution Note: Works consistently up to 1536x1024. Pushing it higher might occasionally generate black borders, though this is rare.</li></ul><h3 id="recommended-settings">Recommended Settings</h3>For the most authentic game feel, I highly recommend using the official Anima base v1.0.<ul><li>LoRA Weight: 1.0</li><li>CFG Scale: 4.5 - 5.0</li><li>Sampler: ER-sde (best for matching the original CG style) or Euler a (for a smoother, more modern look)</li><li>Steps: 30 - 50 (Higher resolutions require more steps to bring out fine details)</li></ul><h3 id="known-issues-(wip)">Known Issues (WIP)</h3><ul><li>Male Character Bleed: The LoRA currently tends to affect the identity of male characters, making them look similar to the game's protagonist, Yasaka Kurou.</li><li>Female characters are largely unaffected and generate well.</li></ul>I will try to fix the male character issue in a future update.<h3 id="training-specs">Training Specs</h3><ul><li>Trainer Tool: <a target="_blank" rel="ugc" href="https://github.com/gazingstars123/Anima-Standalone-Trainer">Anima-Standalone-Trainer</a></li><li>Base Model: Anima base v1.0</li><li>Dataset: 162 lightly cropped CG screenshots (1920x1080). No regularization images used.</li><li>Captioning: Natural language captions via Gemini 3.5 Flash &amp; Qwen3.6 27B-heretic. Danbooru tagging via wd-vit-tagger-v3 (with manual pruning).</li><li>Resolution: Mixed (512 / 1024 / 1536)</li><li>Buckets: Min 512 | Max 1536 | Steps 64</li><li>Batch Size: 4, 4, 2</li><li>Network: Dim 32 | Alpha 32</li><li>Optimizer: AdamW (Weight Decay: 0.01)</li><li>Learning Rate: 0.00002 (Text Encoder LR: 0 — UNet only)</li><li>Scheduler: Cosine (Warmup Steps: 100)</li></ul>

Files with SHA256: 1bc33a21fb307fe7d98a559d824322fd620a1cfdc2e9c2786f38484ca24e6ae2