Based on the portrait style of the video game Stellaris. There are two different versions since the training results varied quite widely. The versions posted are the top performing epoch from the best two training runs. Sample images were generated using the simplest possible prompts just the overall style trigger word and a species trigger.This was my first attempt to put multiple training concepts into a single LORA so some of them don't seem to have taken quite as well as others. I ended up needing to change the trigger word spellings in some cases as using standard spelling was causing issues, some words just couldn't seem to break out during training from whatever the existing meaning for that word in the base model.

Stellaris Portraits LORA

Version B is from a training run with slightly different training settings and time and produces different images to version A. Both versions have their good and bad points.

Files with SHA256: b43def278c4d6cb42df06ca0dfcacc32b361e2f4c4c7447438b6fdff540e21f0