<h3 id="heading-579">2023/6/1 update</h3><p>Update dataset. Enhance training resolution to 1024x1024. Use new craft to retrain the model.</p><h3 id="heading-580">2023/3/31 update</h3><p>Fix the compatibility problem of non-NAI-based checkpoints. Now arbitrary anime model with NAI's VAE or kl-f8-anime2 VAE can also generate good results using this LoRA, theoretically.</p><h3 id="heading-581">2023/3/19 update</h3><p>Add trigger tag '<strong>sakimichan</strong>' to improve generating probability. Reduce training epoch to reduce overfitting effect.</p><h1 id="heading-582">Introduction</h1><p>A LoRA model trained with ~800 images by <a target="_blank" rel="ugc" href="https://www.pixiv.net/users/3384404"><strong>Sakimichan</strong></a>. The 784mb VAEs (NovelAI, Anythingv3, Orangemix, Counterfeit) are recommended. 0.6 ~ 0.8 weights are good. <strong>The preview images are generated using 1.5 version </strong><a target="_blank" rel="ugc" href="https://civarchive.com/models/9139/yesmix"><strong>YesMix.</strong></a></p><h1 id="heading-583">Cautions (Must Read)</h1><ul><li><p>Some preview images are generated using hires fix. Thus, <strong>if you wanna generate the preview images, pls download the original preview images and analyze them using 'PNG INFO' in webui to get all generating information.</strong></p><p></p></li><li><p><strong>The maximum of generating resolution should not exceed 1024 theoretically, otherwise extra/deformed head/limbs will be waiting for you :). </strong>Try to use 'hires fix' to prevent this problem and help to generate higher-resolution images.</p><p></p></li><li><p>For 784mb VAEs (NovelAI, Anythingv3, Orangemix, Counterfeit) users, <strong>make sure add '--no-half-vae' in command line of webui to prevent generating black images.</strong> See <a target="_blank" rel="ugc" href="https://rentry.org/voldy"><strong>voldy's log</strong></a> for the details of edit method. <strong>For high-performance GPU, make sure 'medvram' and 'lowvram' mode are disabled.</strong></p></li></ul>

【Art Style】Sakimichan Style

<p>Fix the compatibility problem of non-NAI-based checkpoints. Now arbitrary anime model with NAI's VAE or kl-f8-anime2 VAE can also generate good results using this LoRA, theoretically.</p>

Files with SHA256: e78faf6c4abadf5c847e46735f93fbba1eda12ee8ee1179f23c314d355494448