Something I created to experiment creating Flux LoRAs. Trained on the official portraits. Can be used to create in-game portraits for your character. Trained images were all 692×1024 so the best results will be creating images of that resolution (or 688×1024 when using Comfy because it refuses 692 for some reason).

I've found that adding realistic helps with genning nice images.

When prompting without any background details you'll simply get a basic colored background. You'll have to specify what type of background you want for it to generate.