<h2 id="overview-ndwl6u3x9">Overview</h2><ul><li>This model is built based on the architecture of the <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294/noobai-xl-nai-xl">NOOBAI XL-VPred 1.0</a>, with some structural modifications. It is trained on the <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/deepghs/danbooru2024">Danbooru2024</a> dataset along with a custom dataset captioned by ChatGPT-4o, and it uses the <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294/noobai-xl-nai-xl">NOOBAI XL-VPred 1.0</a> model as a teacher during training.</li></ul><h2 id="important-note-vv044uxae">Important Note</h2><ul><li>Since this is a model which I reconstructed myself, I would really appreciate any feedback. It will not only motivate me but also help me understand its strengths and weaknesses so I can improve it in the future.</li><li>This is a V-prediction model (different from epsilon-prediction), it requires specific parameter configurations. Please refer to the user guide <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294/noobai-xl-nai-xl">here</a>.</li><li>This model focuses on anime style so it may decrease the performance of artist tag style</li></ul>Currently, the model is not available for use via Civitai Generation. You can visit the following website to use it:<ul><li><a target="_blank" rel="ugc" href="https://tensor.art/models/839278756423152502/IllumiYume-NoobAI-XL-(NAI-XL)-v1.0">tensor.art</a></li></ul><h2 id="recommended-settings-7h4co56rt">Recommended Settings</h2><ul><li>Positive prompt: masterpiece,best quality,amazing quality</li></ul><ul><li>Negative prompt: bad quality,worst quality,worst detail,sketch,censor, simple background,transparent background</li><li>CFG: 4-6</li><li>Clip skip: 2</li><li>Step: 20-30</li><li>Sampler: Euler a</li></ul>Note:<ul><li>I don't use any post-processing and Lora to enhance the example images. I only use these settings and a custom prompt with my base model to generate.</li><li>I used prompts from various sources and authors to generate these example images for comparison and to evaluate my model independently.</li></ul><h2 id="acknowledgments-jc79qce7v">Acknowledgments</h2><ul><li>Thanks to <a target="_blank" rel="ugc" href="https://huggingface.co/narugo">narugo1992</a> and <a target="_blank" rel="ugc" href="https://huggingface.co/nyanko7">Nyanko</a> for sharing such valuable data and <a target="_blank" rel="ugc" href="https://huggingface.co/Laxhar">Laxhar Lab </a>for providing an amazing model!</li></ul><h3 id="if-you'd-like-to-support-my-work-you-can-do-so-through-ko-fi!-4cwna5hyx"> If you'd like to support my work, you can do so through <a target="_blank" rel="ugc" href="https://ko-fi.com/duongve">Ko-fi</a>!</h3>

IllumiYume (Anime Style) NoobAI-XL (NAI-XL)

illumiyumeAnimeStyle_v10.safetensors

v1.0

Overview<ul><li>This model is built based on the architecture of the <a href="https://civitai.com/models/833294/noobai-xl-nai-xl" rel="ugc noopener noreferrer nofollow" target="_blank">NOOBAI XL-VPred 1.0</a>, with some structural modifications. It is trained on the <a href="https://huggingface.co/datasets/deepghs/danbooru2024" rel="ugc noopener noreferrer nofollow" target="_blank">Danbooru2024</a> dataset along with a custom dataset captioned by ChatGPT-4o, and it uses the <a href="https://civitai.com/models/833294/noobai-xl-nai-xl" rel="ugc noopener noreferrer nofollow" target="_blank">NOOBAI XL-VPred 1.0</a> model as a teacher during training.</li></ul>Important Note<ul><li>Since this is a model which I reconstructed myself, I would really appreciate any feedback. It will not only motivate me but also help me understand its strengths and weaknesses so I can improve it in the future.</li><li>This is a V-prediction model (different from epsilon-prediction), it requires specific parameter configurations. Please refer to the user guide <a href="https://civitai.com/models/833294/noobai-xl-nai-xl" rel="ugc noopener noreferrer nofollow" target="_blank">here</a>.</li></ul>Recommended Settings<ul><li>Positive prompt: masterpiece,best quality,amazing quality</li></ul><ul><li>Negative prompt: bad quality,worst quality,worst detail,sketch,censor, simple background,transparent background</li><li>CFG: 4-6</li><li>Clip skip: 2</li><li>Step: 20-30</li><li>Sampler: Euler a</li></ul>Note:<ul><li>I don't use any post-processing and Lora to enhance the example images. I only use these settings and a custom prompt with my base model to generate.</li><li>I used prompts from various sources and authors to generate these example images for comparison and to evaluate my model independently.</li></ul>Acknowledgments<ul><li>Thanks to <a href="https://huggingface.co/narugo" rel="ugc noopener noreferrer nofollow" target="_blank">narugo1992</a> and <a href="https://huggingface.co/nyanko7" rel="ugc noopener noreferrer nofollow" target="_blank">Nyanko</a> for sharing such valuable data and <a href="https://huggingface.co/Laxhar" rel="ugc noopener noreferrer nofollow" target="_blank">Laxhar Lab </a>for providing an amazing model!</li></ul>If you'd like to support my work, you can do so through <a href="https://ko-fi.com/duongve" rel="ugc noopener noreferrer nofollow" target="_blank">Ko-fi</a>!

Files with SHA256: 12bf82fc9526eadd230dd66b39ea01861131382295acea0ae393ecb993aaccc0