First run at creating the ahegao face on realistic faces. Definitely better for i2v. Not much luck with t2v.

I have been running this with 1.0 strength for both low and high noise with better than expected results. I can sometimes get away with no prompt, but if that doesn't work for you try "She does the ahegao face. She sticks her tongue out and crosses her eyes". Still more testing needed.

Optional enhancement to the prompt:

"(Make sure she crosses her eyes during the scene:2)."

Trained on 24 different examples.

Constructive criticism is very welcome. Help me help you.