An embedding for coaxing many goth girls from latent space.
4ygg="Four young goth girls"
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.
Example Images
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:
4ygg - Cons: fairly weak, only hinting. Pros: Doesn't cook your image
4ygg_BoostLt - 4x boost of tensor token values. I generally use this. Fairly reliable with overdoing it.
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.
4ygg_BoostLt_Neg - auto-generated CLIP-perceived antitheses or the input tokens.
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!]