This is my first LoRA, trained on images of the beautiful Christina Ricci in her mid 20s. I've got and learned so much from others here that it would feel wrong not to share the results of my work, even if I'm still a newbie. Took a lot of trial and error to get to this point but I feel like this one deserves to be called v1.0. I've tried it with many of the most popular checkpoints and it seems to hold on pretty well. All the example images are attempts to recreate 1 shot from some of the most popular checkpoints just by injecting the LoRA into the prompt with minimal edits of those original prompts. A single batch of 4 was made for each, I picked my favorite of pic of the 4. No inpainting or photoshop, just the raw first renders.
Like many LoRAs, it seems to work best with weights of 0.6 to 0.8. (used 0.7 in all the test shots, Euler A, 40 steps, CFG 7)
Criticism and advice welcomed, still learning.
edit: added a few example of blonde Christina, there wasn't many example in the source dataset so you need to force SD's hand a little but it's doable. Also added one of my favorite Wednesday creations.
Enjoy!
