<h2 id="an-embedding-for-coaxing-many-goth-girls-from-latent-space.-ktuydlnx1">An embedding for coaxing many goth girls from latent space.</h2><h3 id="4ygg&quot;four-young-goth-girls&quot;-5694whegg">4ygg="Four young goth girls"</h3>This embedding does not harbour any inherent SFW/NSFW biases. It does have an inherent youth bias, so it may be necessary to use appropriate counter-measures if NSFW is your goal. This is very dependent on base model and to a lesser extent, loras used. In my own testing, it can be fairly easy in SFW generation to elicit 'The Shining'-type creepy kids, while in NSFW generations (using base models that don't tend toward wtf!?) I am always presented with young adults.<h3 id="example-images-8q3lqycpy">Example Images</h3>Example images are not cherry-picked, do not use face-restoration, and the prompt contains only the embedding itself. (Exception: negative embedding has a positive prompt of the embedding 4ygg.)Versions:<ul><li>4ygg - Cons: fairly weak, only hinting. Pros: Doesn't cook your image</li><li>4ygg_BoostLt - 4x boost of tensor token values. I generally use this. Fairly reliable with overdoing it.</li><li>4ygg_Boost - 8x boost for very stubborn situations. Don't be surprised by slightly crispy results, especially if values from this embedding intersect with other inputs.</li><li>4ygg_BoostLt_Neg - auto-generated CLIP-perceived antitheses or the input tokens.</li></ul>NOTE: This is an untrained embedding, utilising only text tokens. It does not waste tokens on commas (unlike many similar embeddings posted here and elsewhere).[They're tiny. Download all four. Try 'em out. Then come back here and tell me why you hate them. I dare you!]

Multi Goth Girls

Files with SHA256: 9652d0ebf6f809e6599356f818592bc437ac0dc24c5c0e50e0a1f2a258cb8b75