V3 instructions<ul><li>Syntax is important, ex.1: cum, a naked man, penis. ex.2: cum, a naked latino man with long hair and tattoos, penis</li><li>I use DPM adaptive (important note) for best results with a CFG of 8 and a lora strength of very close to 0.7</li></ul>V2 works much the same as V3 except cum isn't so optional and it was trained on less imagesPlease leave any suggestions on how this model can be improvedsample images upscaled X2 to fix issues with blurry faces<a target="_blank" rel="ugc" href="https://ko-fi.com/alightbulb">https://ko-fi.com/alightbulb</a> if you're feeling generous

Files with SHA256: 502d05f9dfbce30635c37af2c6af1beb0962933beb191d7310e5359e9fdf66cd