Stockings now appear blue and pink much more reliably
The right side of her jacket (from viewer's perspective) is now black much more reliably
Trigger Words
Put (long hair, mismatched stockings:1.15) in positive prompt, and (weapon, sword, holding, holding weapon, holding sword, tail, mechanical tail:1.5) on negative prompt
This is not required but seems to yield better results.
Training Details (compared to Release 3)
The training images are the same, but one image was flipped horizontally, and another was moved to the lower weight folder. Also weight for the higher weight folder increased a little.
Training Parameters
(w/ bmaltais/kohya_ss; see Training Data File for full LoRA training parameters extracted with kohya-ss/sd-webui-additional-networks)
Addresses some attire issues from Release 1~3!!
Improvements (vs. Release 3):
Stockings now appear blue and pink much more reliably
The right side of her jacket (from viewer's perspective) is now black much more reliably
Trigger Words
Put (long hair, mismatched stockings:1.15) in positive prompt, and (weapon, sword, holding, holding weapon, holding sword, tail, mechanical tail:1.5) on negative prompt
This is not required but seems to yield better results.
Training Details (compared to Release 3)
The training images are the same, but one image was flipped horizontally, and another was moved to the lower weight folder. Also weight for the higher weight folder increased a little.
Training Parameters
(w/ bmaltais/kohya_ss; see Training Data File for full LoRA training parameters extracted with kohya-ss/sd-webui-additional-networks)
Trained on: nai.ckpt (925997e9)
Training Batch Size: 2
LR: 0.0007, UNet LR: 0.0007, TE LR: 0.000175, Scheduler: Cosine w/ Restarts
Epochs: 3, Dim: 32, Alpha: 16, Resolution: 512x512, Clip Skip: 2
*EasyNegative in the example images is an embedding from here.