<h1 id="heading-763">Overall</h1><p>This is an stable diffusion lora model, the training dataset with tag words have been placed in training set file. If you would like to use this model just use the safetensors with weight:1 would be ok. Feel Free to use my dataset! <a target="_blank" rel="ugc" href="https://github.com/RuotongX/SD-Nakamura-Yusuke-lora">https://github.com/RuotongX/SD-Nakamura-Yusuke-lora</a> The base model checkpoint is Anylora(LCM), this lora perform best in this checkpoint. The other checkpoint works but not perform good.</p><h2 id="heading-764">Tag</h2><p>You don't need any trigger word, just import this lora would be ok.</p><p>The keywords could be found in my showcases, if you need more you could go to my github and see my training set's tag.<br /></p><h2 id="heading-263">Recommanded checkpoint</h2><p>In version 2, since the base model has been switch to NAI, this lora has pretty decent suitability. But from my test, currently the TMND V checkpoint has the best performance.</p><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/27259?modelVersionId=93612">https://civarchive.com/models/27259?modelVersionId=93612</a></p>

Yojouhan/Asian Kungfu Generation Album cover/Yoru wa Mijikashi/Nakamura Yusuke Styled Lora (四畳半/夜は短し歩けよ乙女/亚洲功夫小子/中村佑介)

<p>Change the base model to Animefull, now this lora has very good suitability to many check points, and logic of generated pictures are better either. However the similarity decrease a little bit, but not too much that still similar to the training style.</p>

Files with SHA256: 0faf042d871625a7463a27c6f9844d701feea4fcb53ff5d170f66b8aa8bdd980