Went to test if I could replicate a specific art project and was shocked at how well it worked.
Trained on a pretty specific photorealistic dataset, but works surprisingly well (and with really surprising flexibility) on animated models (AnythingV5ink worked great).
Starter prompt:
photo of <subject>, rubberface, with rubber bands wrapped around her face distorting her face, <lora:rubberface_v1:1.0>
Advanced prompt (works best on animated models).
<subject>, rubber<target>, with rubber bands wrapped around her <target> distorting her <target>, <lora:rubberface_v1:1.0>