A polished 3D anime style LoRA for Krea 2, trained to reproduce the distinctive visual aesthetic associated with ViciNeko — expressive faces, smooth skin shading, glossy materials, stylized proportions, cinematic lighting, and detailed character rendering.The model was trained with natural-language captions rather than tag-based captions, which helps it respond well to normal descriptive prompts and makes the style easier to control.<h3 id="trigger-words">Trigger words</h3>Primary trigger: <code>vcnk3d</code>Recommended usage: <code>vcnk3d style</code>The trigger can also be used naturally inside a sentence, for example:<pre><code>A cinematic vcnk3d render of an adult woman standing beside a large apartment window at sunset, elegant black outfit, warm rim lighting, detailed hair, polished 3D anime rendering.</code></pre><pre><code>A vcnk3d style full-body portrait of an adult woman with dark purple hair walking through a neon-lit city street at night in the rain, futuristic outfit, wet pavement reflections, purple and cyan cinematic lighting.</code></pre><pre><code>A low-angle vcnk3d-style render of an adult character in a dark fantasy bedroom, saturated purple shadows, glossy materials, detailed face and hair, cinematic lighting.</code></pre><pre><code>A vcnk3d scene featuring an adult anime character standing on a rooftop at sunset, city skyline in the background, warm rim lighting, polished 3D rendering.</code></pre><h3 id="recommended-settings">Recommended settings</h3>- Base model: krea2Anime v1.5- LoRA strength: 0.8- Stronger style: 1.0- Recommended resolution: 1024px+- Prompting style: natural-language descriptionsAt <code>0.8</code>, the LoRA generally keeps a good balance between the base model and the VCNK3D aesthetic.At <code>1.0</code>, the style becomes stronger and more dominant, with heavier stylization of faces, bodies, materials, and lighting.### About the styleVCNK3D is a fan-made style LoRA trained around the visual characteristics associated with ViciNeko's 3D anime artwork.This model is not affiliated with, sponsored by, or endorsed by ViciNeko.

VCNK3D — 3D Anime Style LoRA for Krea 2

- Initial public release of VCNK3D- Cleaned and curated training dataset- Natural-language captions instead of tag-style captions- Improved character and scene descriptions- Trained at 1024 resolution- LoRA rank 32 / alpha 32- 4000-step training run- 3750-step checkpoint selected after side-by-side testing- Tested on krea2Anime, sickOllie and CatTower- Recommended strength: 0.8

Files with SHA256: 95cc430fa9699870f6c1d393ce54738e05d6b66ead0100ef958e515ececabc7c