The goal is for this model to be able to fully understand the nude male. Version 2 is rank 64 so the details are slightly better than version 1. It follows prompts better than prior versions. It is trained on:<ul><li>Nude Male</li><li>Pubic hair (default is closer to a hairless body, you do not need to tag hairless but you can put pubic hair &amp; pubes in the negative)</li><li>Cumming/Ejaculating Cum works well</li><li>Masturbating</li><li>Inserting [object name] in foreskin</li><li>Inserting finger in foreskin</li><li>Grabbing foreskin</li><li>Ass/ Ass hole (example: photo of nude male's rear focusing on the ass hole)</li><li>Very Small/Small penis or Large penis</li><li>Fellatio/inserting tongue in foreskin</li><li>Fellatio/sucking foreskin</li><li>Fully Retracted Foreskin. tag "fully_retracted_foreskin" at the beginning of your prompt</li><li>Partially Retracted Foreskin. tag "partially_retracted_foreskin" at the beginning of your prompt</li><li>Long foreskin - doesn't work as intended.</li><li>Removed Foreskin (I did not train circumcised penis). tag "removed_foreskin" at the beginning of your prompt.</li><li></li></ul>Pros of 2.0:<ul><li>Dataset was tweaked after 1.0 and 1.5 to fill in gaps for some concepts making them stronger</li><li>Drawings/illustrations were recaptioned to avoid overwriting drawing styles as hard. The result are more realistic drawings.</li><li>Trained with less aggressive optimizer parameters making it even more stable than Version 1.0</li><li>It is more consistent with skin texture</li><li>Dataset was tweaked to properly address leaning while sitting so that subjects no longer lean back as frequently</li></ul>Cons of 1.5:<ul><li>Sometimes still ignores prompts</li><li>It was not able to meet many of the goals</li><li>Does not understand females well.</li></ul>Version 1.5 is trained on:<ul><li>Nude Male</li><li>Pubic hair (default is closer to a hairless body, you do not need to tag hairless but you can put pubic hair &amp; pubes in the negative)</li><li>Cumming/Ejaculating Cum works well</li><li>Masturbating - works but the foreskin isn't well defined.</li><li>Inserting [object name] in foreskin</li><li>Inserting finger in foreskin</li><li>Grabbing foreskin</li><li>Ass/ Ass hole (example: photo of nude male's rear focusing on the ass hole)</li><li>Fellatio/inserting tongue in foreskin</li><li>Fellatio/sucking foreskin</li><li>Fully Retracted Foreskin</li><li>Partially Retracted Foreskin</li><li>Long foreskin - doesn't work as intended.</li><li>Removed Foreskin (I did not train circumcised penis)</li><li>Very Small/Small/(default)/Large penis (does not work well)</li></ul>The greens work more consistently. Yellow works sometimes. Red rarely works as intended, but you might get a good shot or two.Pros of 1.5:<ul><li>Dataset was tweaked after 1.0 to fill in gaps for some concepts making them stronger</li><li>Drawings/illustrations were recaptioned to avoid overwriting drawing styles as hard.</li><li>Trained with less aggressive optimizer parameters making it even more stable than Version 1.0</li></ul>Cons of 1.5:<ul><li>It is not consistent with the skin texture. The foreskin texture should be softer.</li><li>subjects frequently lean back or are otherwise not vertical.</li><li>Sometimes still ignores prompts</li></ul>Version 1.0 is is trained on:<ul><li>Nude Male</li><li>Inserting [object name] in foreskin.</li><li>Masturbating</li><li>Cumming/ Ejaculating cum</li><li>Inserting finger in foreskin</li><li>Grabbing foreskin</li><li>Fellatio/ Inserting tongue in foreskin</li><li>Fellatio/ sucking foreskin Going from green to red, the understanding gets worse</li></ul>Cons of 1.0:<ul><li>Drawing styles were overwritten by my illustration style</li></ul>Example Settings:<ul><li>Sampler: Res_2s</li><li>Scheduler: beta57 (download RES4LYF nodes then close fully &amp; restart comfyui)</li><li>Steps - 8-12</li><li>CFG - 1-2</li><li>Shift - 3-5</li><li>Lora Strength - .7-1.0</li></ul><ul><li>Sampler: Euler</li><li>Scheduler: Normal (my examples use "simple" but i think normal gives much better results)</li><li>Steps: 8-12</li><li>Cfg: 1.5 (may go down to 1 for speed)</li><li>Shift: 3.5</li><li>Lora Strength: 1.0</li></ul>The original version prior to 1.0 is just a prototype! I no longer recommend using it because it does not behave consistently.<ul><li>My first version is a prototype. I used ~1200 images but I believe they were poorly captioned, so the resulting images are kind of mid to low adherance to prompts.</li><li>My first version is overtrained. You will need to set the lora weight lower to like .5-.75.</li><li>My first version does not work well with Anime. I decided to test the anime images I tagged for Noob/Illustrious training. It did understand some of the concepts but it adheres to them poorly and has failed to properly capture the anime style.</li></ul>

Zimage Turbo - Foreskin

Files with SHA256: 812096e63a054dc2834a77f5dd445aec140578e03acbb1e3ffb6c7dc352643fd