This LORA needs heavy retraining, it functions well overall - but maybe it needs more guidance in prompts.
It focuses a LOT on the purple/pink lighting.
And while it's 99% realistic data let me warn you: you'll get MUCH better distance on an anime themed or illustration/3d etc model that can take "COLOR THEME" prompting.
This was meant to be entirely DIFFERENT then re:shade as a lot of the data comes from cyberpunk and unsplash data.