Ahegao Face for Wan 14B

(Tested on i2v - not sure if t2v works)

The LoRA pretty much performs how I was hoping, the model will consistently make an ahegao face and drool will drip from their tongue. Due to time constraints my testing was limited. I did 50 videos with the strength at 1 with this prompt:

"A woman is kneeling looking at the viewer. She makes an erotic ahegao face, sticking her tongue out, looking up and crossing her eyes. Drool and spit drip off of her tongue as she sticks it out. "

Unfortunately, sometimes the teeth turn sharp like vampire teeth. In two of my training videos the model had sharpened teeth and it had a stronger impact than I was expecting. I haven't tested if it can be prompted out, but if it is bothering you I would try to add "sharp teeth" and "vampire teeth" to the negative prompt.

The data set was 15 480x854 videos and 9 images. 10 of the videos were in portrait orientation, 5 in landscape.