Shoop Da Woop (I'M A' FIRIN' MAH LAZER!!) - XL BETA API Documentation

So Ive tried training it in XL at 256 resolution (which does indeed seem to work), but I keep overtraining it. So far, this is the best version Ive gotten. It seems a bit inconsistent with the upscaler, but produces good images without need to be upscaled sometimes. It may depend on the model. Either way, it seems to work best around 0.75 strength and using kewyord style prompts. "open mouth" and "red lips" seem to help get the LORA to work better on some subjects, but they arent always necessary