<h1 id="luminelle">✨ Luminelle ✨</h1>Luminelle is my attempt to create a 2.5D model that combines the detail and depth of realism with the aesthetics and visual style of anime.I plan to gradually integrate my own LoRAs into the model as they are released, with the goal of further improving its quality and capabilities.The model still needs more testing, so any feedback, examples, and suggestions are greatly appreciated! 💙<h2 id="version-1.0">Version 1.0</h2>This is the base for future improvements, built from a merge of several models and my own LoRAs. It serves as the starting point for future versions.<h2 id="version-2.0">Version 2.0</h2><ul><li>In V1, the colour palette leaned heavily towards warm tones. Colors should now appear more neutral. If you prefer the warmer colour palette from V1, use the <code>warm colors</code> tag.</li><li>Improved overall detail, particularly in the background.</li><li>The model leans slightly more towards realism while still trying to preserve the original style.</li></ul><h3 id="recommended-settings">Recommended Settings ⚙️</h3><ul><li>Steps: 25–35</li><li>Sampler: Euler a</li><li>CFG Scale: 5–7</li><li>VAE: Baked-in</li></ul>Recommended Positive Prompt: <code>masterpiece, best quality, absurdres</code>Recommended Negative Prompt: <code>lowres, bad anatomy, bad hands, worst quality, low quality</code>Feel free to experiment with the settings - different prompts and LoRAs may benefit from slightly different values.<h3 id="limitations">Limitations</h3>Like most SDXL-based models, Luminelle can still struggle with very small objects and fine details, which may result in malformed or inconsistent elements. ADetailer can help reduce these issues, especially for faces and other important details.<h3 id="recommended-workflow"><a target="_blank" rel="ugc" href="https://civarchive.com/models/1386234/comfyui-image-workflows?modelVersionId=3265669">Recommended workflow</a></h3>

Files with SHA256: 95321bd610aa25b554799fe339ff6b4656141c35c5e4734e462486e472ddfa17