I´ve made this textual inversion, the first one i believe is "good"! it´s based on some images from instagram from the influencer @draculangelica as one of her characters "trixy treats"! It´s a halloween themed character and she´s super sexy :Dhere you can find some pictures of her:<a target="_blank" rel="ugc" href="https://www.google.com/search?q=trixy+treats&amp;client=opera-gx&amp;hs=pyh&amp;sxsrf=AJOqlzUM_Ef676zIFMBe4l58ZB-oPnFEIg:1675599031973&amp;source=lnms&amp;tbm=isch&amp;sa=X&amp;ved=2ahUKEwj7wNjorP78AhUSX_EDHTV7CH0Q_AUoAXoECAEQAw&amp;biw=1879&amp;bih=931&amp;dpr=1">https://www.google.com/search?q=trixy+treats&amp;client=opera-gx&amp;hs=pyh&amp;sxsrf=AJOqlzUM_Ef676zIFMBe4l58ZB-oPnFEIg:1675599031973&amp;source=lnms&amp;tbm=isch&amp;sa=X&amp;ved=2ahUKEwj7wNjorP78AhUSX_EDHTV7CH0Q_AUoAXoECAEQAw&amp;biw=1879&amp;bih=931&amp;dpr=1</a>-&gt; important &lt;-i figured out, that adding "room, no person, empty room, architectural" prevents the model to generate just some room with the same color scheme :D-&gt; important &lt;-the sample grids have been made with different models available here on civitai, like protogenX53, eldrethsOG4060 and others, for the model to work, just put "trixy_dracul" in your prompt, found it to best work at the beginning! adding ((((beautiful face)))) ((perfect face)) seemed to give better results, without there was some face distortion ...the training images where full/upper body shots and only the background was changed to be different on all images, so the whole look would somehow be included in the modelthe facial makeup is a bit different, SD produces sometimes a bit strange stuff here, but it´s okay i think :Denjoy

Files with SHA256: 3c1690a01b3dbe78bd24dc21d17cf22ec48e21dbaeb643837ca63ea22d9ce19a