<p>Generates "Boykisser" from that one meme.</p><p></p><h2 id="tested-on">Tested on</h2><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/2754368/anima-int8-mxfp8?modelVersionId=3099040">Anima Base v1.0 int8</a></p><h2 id="showcase-images-settings:">Showcase images settings:</h2><p>Sampler: LCM (other samplers should also work fine)</p><p>Sched: sgm_uniform</p><p>CFG: 1.0/2.0</p><p>Resolution: 512x768 -- Upscaled 4x with <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/2x-StarSample-V2">StarSample V2</a>.</p><p></p><h2 id="recommended-settings">Recommended Settings</h2><ul><li><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/2560840/anima-turbo-lora">Anima Turbo LoRA</a> (for better style accuracy)</p></li><li><p>LoRA strength ~0.8 (the lora starts to fry at &gt;=1.0)</p></li></ul><h2 id="training-configuration">Training Configuration</h2><p>Steps: 300</p><p>LR: 1e-4</p><p>Optimizer: AdamW8bit</p><p>TE: frozen</p><p>images: 10</p><p>resolution: ~480²</p><p>Duration: 25 min - 4min dataset; 21 min training (~4.17s/it on tu116)</p>

Files with SHA256: bf32f75bc02c5aa652031a85192ea1997a6e3d614b74346dc32a3ca1fe99fe24