Generates a likeness of the streamer.
This LORA prefers a lower guidance value (3-6).
Let me know if it is over/undertrained.
Description
new version is trained on more photos at a lower rate
FAQ
Comments (2)
This new version looks way better. Nice job. Out of curiosity, how many photos did you start with and how many more is this one?
same setup as before, but I trained it at a lower learning rate for much longer