Character/style (SAGE)

Trained on SDXL 1.0, this LoRA is built from a curated collection of AI-generated images of freckled women, set against cinematic, dynamically lit backgrounds.

With this LoRA active, you’ll likely see better depth and shading overall. Faces tend to have richer, more dynamic lighting, and skin looks more natural and textured, almost like it's alive. Freckles show up in a more realistic way, and you get subtle skin imperfections that make everything feel more believable.

One of the best parts is the consistency it brings to characters. The portraits usually stick to a similar aesthetic. I noticed the lighting, tones, and little details often match across images, almost like they were shot during the same session. In testing, it also helped fix flat-looking portraits and added more life and variation to the output.