<h1 id="soft-dramatic-lighting-lora-52nkjz2ln">Soft Dramatic Lighting LoRA</h1>Style LoRA that is trained shadow, contrast, lighting, and darknessMy ComfyUI Workflow: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/14582">Comfyui Workflow: SDXL FaceDetailer, Upscaler, &amp; Batch Selection</a>Pony Version 2Strength: 0.1 to 0.9 is the range. Best is between 0.5 to 0.75. Posted images are 0.60Pony Version 1Strength: 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.<h3 id="tips-yfsrysl4m">Tips</h3>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.<h3 id="training-ohsyxidp9">Training</h3>All images were tagged using WD14.

Soft Dramatic Lighting

10+ more images to the dataset with more variety. Adjusted learning rate (Prodigy) and turned off Text Encoder (TE1)Adjusted tags for better results

Files with SHA256: 19b8ffd7b5a943f021d797e6ed8acda3d2e0afc149a4647c6fdf4ed5a58bacbc