This embedding was trained on 14 images taken from model cards from the original versions of other models of mine.
It seems to want to add extra details and just zhuzhes things up a bit - see example images for comparisons with and without the embedding from a variety of models/scenes
Use the filename to trigger the embedding (so should be rcnz_style_v3 unless you rename it)
Big thanks to Juice and Airic for helping test this out, and ensuring it's worth using!