Soft Dramatic Lighting LoRA
Style LoRA that is trained shadow, contrast, lighting, and darkness
My ComfyUI Workflow: Comfyui Workflow: SDXL FaceDetailer, Upscaler, & Batch Selection
Pony Version 2
Strength: 0.1 to 0.9 is the range. Best is between 0.5 to 0.75. Posted images are 0.60
Pony Version 1
Strength: 0.1 to 0.9 is the range. 0.7 to 0.85 is probably good. 0.9 to 1.0 may cause burning issues.
Tips
May need to prompt "looking at view" because a lot of the training sets for this LoRA were looking away from the camera.
May need to use "blurry" or "blurry foreground" as negative prompt. Lots of the training images had different focal points may cause blurry generations around the subject.
Avoid prompts like "light", "dim light" in the positive prompt section. The generator will add light to the scene.
Training
All images were tagged using WD14.
Description
10+ more images to the dataset with more variety.
Adjusted learning rate (Prodigy) and turned off Text Encoder (TE1)
Adjusted tags for better results