If you've used my SD or illustrious versions of this, you know the drill. Does what it says on the tin. Best at views from behind but can do from the front, especially with image to image. Best used by generating the base image, painting the mess on, then regenerating with the LoRA, though it will work decently well with just a good prompt. Issue is the automatic prompt helper tends to censor certain details on the mess.
I did some minor tweaks to the training data from the other versions so this took a lot of training credits, but it's worth it because DiT is great. If you're reading this you probably use this site for a similar purpose which LoRas make so much better. Enjoy
