<p>Style based on Akasaki's <a rel="ugc" href="https://www.youtube.com/watch?v=RCltAg_iK0E">Bunny Girl</a> MV. I extracted some key frames from the video. Since there are basically just two characters in the video, I trained a base model to help me draw new versions of the same frames, introducing character variety. This model was trained with the manually-extended dataset. As usual, I trained on WAI v12.0 as checkpoint.</p><p>The trigger word is <span style="color:rgb(250, 82, 82)">akasaki_bunny_girl_style</span>. The LORA weight works fine at 1.0. You may reduce the trigger word's weight for specific characters. It has worked fine between 0.8 and 1.0 for me; take a look at the sample images.</p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/RCltAg_iK0E?rel=1" start="0"></iframe></div><p></p>

Files with SHA256: 548afd255a2dacb4cb8e2a47740aa4e54ae65fe391434229aec6b1967b167e40