Type
LORA
Publish Time
2024-12-07
Base Model
Flux.1 D
Trigger Words
woman
Copy
Version Introduction
I used 4 images of high quality and did 4 epochs of 800 steps. This was a test but it turned out pretty good. I didn't even know who Emily was but just saw her picture when I was looking for someone else. She's definitely photogenic. :) The trick is you need at least one picture of the subject smiling, one neutral, one looking to the side, and one making eye contact. If you don't, the trained LoRA will either not be able to look straight, be stuck with only a smile or be unable to smile. It makes sense but it isn't easy to find pictures where people aren't smiling. :) See? I just smiled.
