Riley Fairfeather character lora
V1 - 1575 steps - Trained on SD 1.5 EMA
V2 - 3150 steps - Trained on SD 1.5 EMA
V3 - 2625 steps - 1st of 3 epochs which turned out best, Trained on SD 1.5 EMA
V4 - 5512 steps - 3rd of 3 epochs, all were very similar, Trained on AOM3
Description
2625 steps trained on SD 1.5, 250 steps per image, 21 images.
The version I'm posting is Epoch 1, as both Epoch 2 and Epoch 3 have issues where they create duplicates/clones.
FAQ
Comments (1)
Just as a comment, V3 and V4 were released at the same time. The primary difference between v3 and v4 is the training base model used.