Modify the weights of the LoRA to achieve the desired result. The base model for SD 1.5 will look the best. However I had great results with Protogen, Dreamlike Photoreal, and Realistic Vision as well, and various blends.
This was also trained with the expanded VAE located here. Please use this VAE when inferring with this LoRA.
Description
Used bmaltais's implementation of kohya_ss LoRA training scripts.
Resumed training on the original 1.0 model with a bit more focus on full body photos. AOC at a distance will still require inpainting of the face. Training was resumed at 768x768 but the best generations will likely be 512x786 (portrait).
The sample images were generated with the base SD 1.5 model.
Total Images: 19 total high resolution images from the original dataset in version 1.0
Regularization Images: 300 total "sexy woman" regularization images from ProGamerGov's dataset at huggingface.
Trigger word: alexandriaocasiocortez woman
Training Steps: 2100 in 1.1, 4000 total including 1.0
Unet Training Rate: 1e-04 (0.0001)
Text Encoder Training Rate: 5.0e-05 (0.00005)