This LoRA is designed for a semi-realistic Niji style. It was trained on images from Niji Journey, with 90% of the dataset focused on male characters.For warm color tones, use tag: “warm theme” (highly recomended) For cold and blue tones, use tag: “cold theme”To achieve a more realistic look, use the following tags:<pre><code>(realism, photorealism, cinematic lighting, volumetric lighting:1.2)</code></pre>Recommended strength: 0.6-0.8 Lower CFG values produce better results, recommended CFG 1–5Tip: if the result looks oversaturated or too glossy, simply lower the CFG scale, this usually gives a much better result. It’s also best to avoid checkpoints that have a strong built-in glossy-AI style.<h3 id="i-recommend-using-this-lora-with-anime-checkpoints.">I recommend using this LoRA with anime checkpoints.</h3>The best results are achieved when paired with NOOBAI checkpoints, since they have a much better understanding of painterly textures.In contrast, Illustrious tends to suffer from a plasticky look.<s>I do NOT recommend</s><s> using this LoRA with bara checkpoints (homosimile, xavier, etc.), as they can heavily distort facial structure and produce strange features due to the baked-in bara style.</s><h3 id="v2:">V2:</h3><ul><li>Stronger style effect</li><li>Improved dataset</li><li>Network DIM increased to 96 (potentially better textures)</li></ul>I recommend using this LoRA with the models listed below.<a target="_blank" rel="ugc" href="https://tensor.art/models/947801774926900128?source_id=kT61oVzqkkm6o_ooYH7y8hMj">Tensor.art link</a>

Niji men XL

nijimenV2.safetensors

v2.0

v1.0

This LoRA is designed for a semi-realistic Niji style. It was trained on images from Niji Journey (around 170 images), with 90% of the dataset focused on male characters. (but it can be used to generate female characters)For warm color tones, use tag: “warm theme” (highly recomended) For cold and blue tones, use tag: “cold theme”To achieve a more realistic look, use the following tags:(realism, photorealism, cinematic lighting, volumetric lighting:1.2)Recommended strength: 0.6-0.8.Lower CFG values produce better results, recommended CFG 1–5All images on the showcase were generated in ReForge webUI, using the base model Chenkin Noob XL (CKXL), with no post-processing in other apps. (Just click “remix” on any image from the showcase and select the <a target="_blank" rel="noopener noreferrer nofollow" href="https://tensor.art/models/946606575722727661">Chenkin v0.2</a> base model to get the same result.)Tip: if the result looks oversaturated or too glossy, simply lower the CFG scale, this usually gives a much better result. It’s also best to avoid checkpoints that have a strong built-in glossy-AI style.I recommend using this LoRA with anime checkpoints.The best results are achieved when paired with NOOBAI checkpoints, since they have a much better understanding of painterly textures.In contrast, Illustrious tends to suffer from a plasticky look.<s>I do NOT recommend</s><s> using this LoRA with bara checkpoints (homosimile, xavier, etc.), as they can heavily distort facial structure and produce strange features due to the baked-in bara style.</s>V2:<ul><li>Stronger style effect</li><li>Improved dataset</li><li>Network DIM increased to 96 (potentially better textures)</li></ul><a target="_blank" rel="noopener noreferrer nofollow" href="https://civitai.com/models/2193733">CivatAI</a>

Files with SHA256: 7ca9687985ae5332965e76bdfa35e62d58083dd662efb87bd33fbbd8b02e46a8