This might be a little iffy since it's my first LORA. I noticed that there weren't too many overexaggerated expression LORAs for Illustrious, so I decided to try it out myself. The training data was specifically from "The Mask" animated TV series. The focus was originally on the "cartoon eye-pop" that AI has trouble doing, but it seems to have created an overexaggerated cartoonish expression.

Tested on Realisticnoobmix, HomosimileXL, HomosimileNAI version, Homosimileveritas, and waiNSFWIllustrious with decent results.

I'm really curious how others' outputs turn out, so please post pics if you get the chance! Feel free to copy my prompting for help. I often have a lot of wording in the negative. I also sometimes use ADetailer too.

Prompting Help

Possible Triggers for the Positive: maskexag, extremely wide open mouth, long tongue, large bulging eyes stretched out of face with red veins, excited, surprised, shocked,

Red veins may be tricky because it will sometimes put them on the body and the face instead. You can exclude them if you don't want them, or even just put in the negative.

Possible Negatives:

-green face (to avoid green-masked/bald characters)

-bald (might want to describe the hair in the Positive to help)

-bunny ears (AI may have trouble with the stretched eyes and often misinterprets, I noticed this when using Joycaption for the images)

-red circlets around eye sockets

Strength: 1 should be fine, but it really depends on what you're looking for. You may need to enhance certain words/phrases with parentheses, but make sure to reuse the same seed to see if there's much variation.

Here is an example of the difference with/without the LORA:

Prompt was: side view, realistic, masterpiece, best quality, newest, perfect quality, detailed background, gym, 1boy, male, long bulging eyes stretching out of face with red veins, grin, extremely excited, short hair, brown hair, wearing black tank top, wearing blue sweatpants, running, thin

It may be best to help strengthen the output with another Mask LORA if you are trying to mask a character. Again, you may have to adjust the strength.

These are my usual go-to Mask LORAs that work well:

She-Mask (will do male/female easily): https://civarchive.com/models/1048485/the-mask-or-she-mask-omni-mix

Animated The Mask Stanley: https://civarchive.com/models/1680332/stanley-ipkiss-mask-the-mask-commission

Masked Milo and Otis: https://civarchive.com/models/1648991/milo-and-otis-the-mask

Masked Dorian: https://civarchive.com/models/1132494/the-mask-dorian

Masked Tim: https://civarchive.com/models/1071282/the-mask-tim

I used captions for the training data images. However, after refining and finishing it, Civitai then told me afterwards that tags are better for Illustrious.... 🙃

I'll have to look into trying this again with tags to see if there's much of a difference.