This was trained three times using the method outlined here. Consequently this LoRA requires a strength pf 0.1 or maximum 0.12. More than that, and you will get garbage results.
An entirely new dataset was created for this LoRA, and it is far more flexible than V1.0. However, due to the great difficulty in posting images of real people to Civit, I am only providing one face picture. You will have to try it for yourself.
If you prepend the word 'young', the hairstyle may change from Barbeau's classic curly perm of 1979-185, into a straighter style from the early 1970s.
This was trained three times using the method outlined here. Consequently this LoRA requires a strength pf 0.1 or maximum 0.12. More than that, and you will get garbage results.
An entirely new dataset was created for this LoRA, and it is far more flexible than V1.0. However, due to the great difficulty in posting images of real people to Civit, I am only providing one face picture. You will have to try it for yourself.
If you prepend the word 'young', the hairstyle may change from Barbeau's classic curly perm of 1979-185, into a straighter style from the early 1970s.