The Sanderson Sisters LoRa for Pony XL

Trained on movie stills featuring all three characters in the same frame.

This was mostly an experiment to see how well multiple characters can be trained into the same model. It works with varied results, but I would still say it is pretty successful.

Seems to do best with landscape generation. Try using higher CFG and Steps as well as additional LoRas like Detail Tweaker and Detailed Eyes.

Could work really well if you are able to upscale.