<h3 id="heading-69">2023/7/17 update</h3>Upload the realistic version to generate images using arbitrary realistic models. Please make sure to set the clip skip as 2 when generating images. The generalization ability of the realistic version has been tested on multiple realistic models. See the preview images for details.<h3 id="heading-227">2023/7/16 update</h3>Update dataset. Retrain the model using new craft.<h3 id="heading-228">2023/3/11 update</h3>Update dataset and model. Reduce loss to 0.09 using Lion optimizer. If you want to stably generate good results, pls make sure your sampler settings are identical to the following settings (and make sure your clip_skip is set as 2):<ul><li>eta for DDIM: 1 (0.67~1 should be ok)</li><li>eta for ancestral samplers: 1 (must be set as 1)</li><li>sigma noise: 1</li><li>eta seed noise delta: 31337</li></ul><h1 id="heading-229">Introduction</h1>A LORA model trained with ~28 images of Shiranui Mai by <a target="_blank" rel="ugc" href="https://twitter.com/Akiryooooo">Akiryo</a>. The trigger tags are 'akiryo-mai' and 'shiranui mai'. 0.6~0.8 weights are good. The preview images are generated using latest version of <a target="_blank" rel="ugc" href="https://civarchive.com/models/9139/yesmix">YesMix </a>and <a target="_blank" rel="ugc" href="https://civarchive.com/models/7808/easynegative">EasyNegative </a>embedding.

【Character】Akiryo's Mai

Files with SHA256: 38c110908ba4e08b29c1c62a4b035319afc33278cdd3317e1f22156da3bd1b08