<h1 id="raehoshi-anima">Raehoshi Anima</h1>Raehoshi Anima is an enhanced iteration built upon the <a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima">Anima Base v1.0</a> architecture. This release focuses on elevating visual style, integrating extensive new concepts, and expanding character knowledge. The ultimate goal is to deliver a more polished, balanced, and visually stunning output while remaining faithful to the core strengths of the base model.<h2 id="installation-and-requirements">Installation &amp; Requirements</h2>Important Note: This model does not include a built-in Text Encoder or VAE. You must download these components separately to achieve the intended results.File Placement Guide<ul><li>raehoshi-anima-v1.0.safetensors goes in ComfyUI/models/diffusion_models</li><li>qwen_3_06b_base.safetensors goes in ComfyUI/models/text_encoders</li><li>qwen_image_vae.safetensors goes in ComfyUI/models/vae</li></ul><h2 id="recommended-settings">Recommended Settings</h2>For the optimal experience and the highest quality generations, we recommend the following configurations:<ul><li>Sampler: <code>Euler a</code> or <code>ER SDE</code></li><li>Schedule Type: <code>Beta</code> or <code>Normal</code></li><li>Steps: 32</li><li>CFG Scale: 4.0 – 5.0</li><li>Resolution: Any resolution up to 1536 (ensure dimensions are divisible by 32)</li><li>Positive Prompt: <code>masterpiece, best quality, score_7, absurdres</code></li><li>Negative Prompt: <code>worst quality, low quality, score_1, score_2, score_3, artist name, blurry, jpeg artifacts, bad anatomy, bad hands, bad proportions, mutation, deformed, extra digits, fewer digits, missing arms, missing legs</code></li></ul><h2 id="prompting-tips">Prompting Tips</h2><ul><li>Tag Ordering: For the most consistent results, follow this structured prompt order:<code>[Quality / Meta / Year / Safety tags]</code> ➔ <code>[1girl / 1boy / Character Count]</code> ➔ <code>[Character Name]</code> ➔ <code>[Series / Copyright]</code> ➔ <code>[Artist]</code> ➔ <code>[General Tags]</code></li><li>Character Accuracy: Always include the official series/copyright tags alongside the character name to significantly improve details and accuracy.</li><li>Hybrid Prompting: The model handles hybrid prompting seamlessly. Feel free to mix dan match danbooru-style tags with natural language descriptions (e.g., use tags for characters and natural language for background/action).</li></ul><h2 id="training-details">Training Details</h2>Raehoshi Anima was trained using a <a target="_blank" rel="ugc" href="https://github.com/Raelina-Rae/diffusion-pipe">custom personal fork of Diffusion-pipe</a> across a comprehensive two-stage fine-tuning process. The dataset utilizes multi-level captioning with random selection and tag dropout to ensure flexibility.<h3 id="stage-1:-concept-and-character-expansion">Stage 1: Concept &amp; Character Expansion</h3><ul><li>Dataset Size: ~25k images</li><li>Trained Resolution: 1024x1024</li><li>Hardware: NVIDIA RTX PRO 6000 (96GB VRAM)</li><li>Batch Size: 32</li><li>Gradient accumulation steps: 1</li><li>Learning Rate: <code>1.5e-6</code> (LLM Adapter LR: <code>2e-7</code>)</li><li>Focus: Introducing new franchises, series, and character knowledge.</li></ul><h3 id="stage-2:-aesthetic-and-style-refinement">Stage 2: Aesthetic &amp; Style Refinement</h3><ul><li>Dataset Size: ~1k high-curation images</li><li>Trained Resolution: Multi-aspect (1024x1024 &amp; 1536x1536)</li><li>Hardware: NVIDIA RTX PRO 6000 (96GB VRAM)</li><li>Batch Size: Per-resolution batch size (24-1536x1536) &amp; (48-1024x1024)</li><li>Gradient accumulation steps: 1</li><li>Learning Rate: <code>1e-6</code> (LLM Adapter LR: <code>0</code>)</li><li>Focus: Mitigating artifacts, balancing composition, and enhancing the overall visual style.</li></ul><h3 id="list-of-new-seriescharacters-trained:">List of New Series/Characters Trained:</h3>Expanded Knowledge Base (Up to May 2026)<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">The model’s character and lore library has been updated to include the latest data for:<ul><li>Zenless Zone Zero</li><li>Wuthering Waves</li><li>Honkai: Star Rail</li><li>Genshin Impact</li><li>Arknights: Endfield</li><li>Neverness to Everness</li></ul><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">For character trait details prompts, please refer to the Danbooru site for accurate tags and references.<h3 id="special-thanks">Special Thanks</h3>A huge thank you to GSlinux for providing the development support needed to make this project a reality.<h3 id="support-the-development">Support the Development</h3>If you love using this model and want to help fund future iterations and dataset curation, consider supporting the project:<ul><li>⚡ Send a tip of Yellow Buzz directly on this platform.</li><li>☕ Buy me a coffee via <a target="_blank" rel="ugc" href="https://ko-fi.com/raelina">Ko-fi</a></li></ul><h2 id="license">License</h2>This model is released under the <a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima/blob/main/LICENSE.md">CircleStone Labs Non-Commercial License</a>.

Files with SHA256: 50035174a66fca476700eba6c85a3e8e3e284f8c5d9026954402fbd4b1306463