For FLux: use 0.9-1 strength
V1.5> use 0.7 strength
SDXL:
Trained on 100 photos from Stefan Gesell.
Mixed 3 different epochs.
Usually works fine with a weight of 0.5-0.9
Description
hat to completely redo the data set but here is the first version with the smaller pictures and only 24 pictures.