EDIT: THIS IS AN ENTIRE CHECKPOINT MODEL DEDICATED TO THIS CHARACTER NOT A LORA. ADD YOUR STYLIZED LORAS AFTER LOADING THE MODEL, IT WORKS BETTER THAT WAY. THIS WAS TRAINED WITH DREAMBOOTH.This is Dr. Beverly Crusher from Star Trek the next generation. You can run this model at 512x512 (recommended) or 768x768. I would recommend using Adetailer extension for cleaning up more distant face and hand shots. This model was Dreambooth trained with Koyha-Gui on 128 images, closeups, distant shots, merged with other models, then re-merged again at very low strength 1.0 with one of the "overtrained" checkpoints Kohya spit out at the end. If you have questions on the worldflow please feel free to ask. Dreambooth definitely gives better results than SD 1.5 Loras and SDXL Loras for subjects.I recommend using --no-half --no-half-vae for startup argument in auto1111 for higher quality, and the VAE vae-ft-mse-840000-ema-pruned.safetensors for this, adding "weird eyes" to the negative prompt. You can also use terms such as "full body shot" for a more distant view, "rear view" "side view" "closeup"

Files with SHA256: 91a6d965b285effb6d518ffd29a7f8c30a0a2f331782a6bc9fb4dee5afcff93b