My third model. Honestly, I do not know for who it can be useful, because it is literally a hole from which the tongue sticks out. But maybe it's just my lack of imagination :)

v1.0
Using some good prompts, u can get really incredible results. For example, in dataset of model zero images with eyes of girl, only one tongue, but if u simply prompt "eyes", it's significant increase chance, that picture will contain whole face of girl. Also this model works well with my another model "Open Gag Mouth"

Tips: 0.6 - 0.8 scale will be good (with some LoRas need higher or lower)
Training model (also used to generation): abyssorangemix2_Hard
Keywords: glory hole tongue, pov, cum on tongue, salva on tongue, open gag mouth (last tag maybe bad working, so pls for this option download my another model Open Gag Mouth)

v2.0
Using this LoRA, you get an image with the face of a girl who sticks out her tongue. By adding the prompt "cum on tongue" you add cum to the tongue and some parts of the glory hole (not always). Works with character LoRA's (not good with all, for some LoRa's u need to lower scale), the example shows LoRA Lucy from Edgerunners and Prinz Eugene from Azur Lane. Maybe it's a little overbaked, but what do you want from a hole with a tongue in it? It was also quite difficult to assemble the dataset, even the one on which this LoRA model trained.

Tips: 1 scale will be good (with some Character LoRas need higher or lower)
Training model: NAI
Model used to generation: abyssorangemix2_Hard
Keywords: glory hole tongue (also u can add "cum on tongue")

If u know how to improve model, let me know.