@Bendy and Boney
The image was indeed created with “Yiff in Hell”. I’ll pay attention to that in the future and adjust the tags in the prompt accordingly.
I’m still practicing, and if I like what I see, then I don’t mind things like the lettering. Especially when I think about what else gets printed on clothes.
the pony diffusion v6 model was trained with score_[number]_up (which decile the image was in the source sites’ score ranking, iirc) in its dataset as a way to prompt for quality, but the image looks like it was made with a non-pony diffusion v6 sdxl model, where the score tags aren’t associated with anything.
i’ve found with those models that if you put something in the prompt that the dataset doesn’t associate with a tag, it sometimes attempts to put it in as writing somewhere in the image (key word here is ‘attempts’).
The image was indeed created with “Yiff in Hell”. I’ll pay attention to that in the future and adjust the tags in the prompt accordingly.
I’m still practicing, and if I like what I see, then I don’t mind things like the lettering. Especially when I think about what else gets printed on clothes.
score_[number]_up(which decile the image was in the source sites’ score ranking, iirc) in its dataset as a way to prompt for quality, but the image looks like it was made with a non-pony diffusion v6 sdxl model, where thescoretags aren’t associated with anything.