This my first try to do a textual inversion. It's suppose to display women showing cum in their mouths. The textual inversion seems to work on different checkpoint models but you probably need to try different prompts before getting an ok image.
It is trained on 26 images of women of different background but example images are of blondes.
Don't use "Restore faces".
Description
FAQ
Comments (9)
safetensor please?
Not sure how to make safetensor when training a textual inversion in AUTOMATIC1111/stable-diffusion-webui. Tips or links? But I'll try to make safetensor next time if possible.
This is an embedding... I dont get the demand for a checkpoint format. Either I'm missing something or folks are dumbos
@Crept Safetensor is safer & Im obviously asking for a Lora version since TIs dont have the popular safetensor
It would be good. Easy Diffusion has no way to use textual inversion or I'm too newbie to figure out how.
My SD won't recognise this as a TI...
Good for inpainting, bad for txt2img.
Got fairly poor results when using this in txt2img, both with realistic (stablydiffusedsMagnum_v10 and the model OP used in his examples realisticVisionV13_v13) and anime-ish models (sxzLuma and crazyGirlAshleyMix2_20). However, it performs well when inpainting, and this is something that should have been mentioned in the instructions - I would have given it a one-star rating for not working if I hadn't tried it myself and only attempted txt2img like in the examples.
The example images here were generated with the same model used in OP's example images, using a slightly modified version of OP's example prompt. You can see that the txt2img images don't show the expected result, but that's easily fixed with inpainting with a very simple cum_in_mouth, white cum in mouth prompt after drawing a mask over the mouth. And it definitely does have an impact - removing it from the inpainting prompt shows significantly poorer results, as you can see in the images I've posted in the gallery.
Thanks I was using the sample photos through PNGinfo and getting terrible results
Struggling to generate a good result with this. I haven't fully isolated it in my mess, but it seems to add a strong bias towards deep-purple tongues, top-down perspective and an overall indigo tint that is very difficult to offset.
As soon as I remove it: the tint is gone, the tongues are red as they are supposed to be and it's not locked to a top-down shot anymore.