Behold! Arguably my most generic LoRa yet! As there are already a butt-jillion other LoRas with this exact same concept. And truth be told, I love pretty much all of them. However, I have a gajillion blue buzz, a very small dataset, and an itch for experimentation. Lo and behold, figured I'd try it myself. We like 'em beeg.
Should play nicely with both anime and realistic images. Although you will want "realistic" in negatives if you're doing any 2d style.
Description
Trained onsite. Dataset 5 images.











