About:

This Textual Inversion is based on Megan Fox post-surgery.

Version Notes:

Trained with 21 photos. Stopped training at 2,400 steps due to overtraining.

Usage Tips:

I've found that this works best when Restore Faces is not checked. Also try to add 1.1 preference to trigger word, for example:

(Megan_Fox_512-2400:1.1)

This also works well with ADetailer extension. Use model face_yolov8n.pt and prompts:

Megan_Fox_512-2400