This is for women with their tongues out. It has no eye rolling and is biased towards eye contact or having eyes closed. It is biased towards closeups but can be made to do other things easily. It can be made to push mouths open much wider than ahegao lora or the popular awaitingtongue embedding. It does well with side front or top shots. It has a tendency for split tongues especially when upscaling, reduce the cfg scale, and reduce the weight on 'tongue out'

This is a LOHA trained with 16/8/4/2, lr 2e-5 and 0.8e-5 for22 epochs on a set of 50 or so carefully curated super-close cropped images, all with identical tags "1 girl, mouth open, tongue out, extreme closeup"

You need both activation tags to make it work but you can play with the strength. It works well with any model.

Future versions may offer an ahegao option.