Makes things more realistic looking, works better for text2image than text2video but works fine on both, just add it as a lora, no trigger words.