m0uth is the trigger for the Lora, past that add into your description parts that you want to see (ie: open mouth, tongue, tongue_out, uvula, tonsils, pharynx).
This LoRA is trained on images of the detailed insides of a woman's mouth. This should improve the realism and prevent incorrect oral anatomy especially on close-ups. The training also involed the pharynx (the back of the throat behind the uvula).
Description
There seems to be a trend of having a second uvula further down the throat on more close up images, I am not sure what is causing this, possibly shadows in the training data. I will try to remedy this in future versions.
This version has a tendency of doing mouth only images, especially if you use more mouth based descriptions.