Works quite well to guide the style of pony-based models towards a more painterly semirealistic style. As is with many style loras, the resulting style is not really wlop, but there is still a wlop-like vibe to the resulting images.

Trained with 150 images from wlop which were captioned using gpt-4-vision in an attempt to make more responsive to natural language. It's anyone's guess whether it worked or not, but the result is pretty good on any pony-variant models.

Note: horns aren't really part of the lora, I was just into horns at the moment of making the examples ;)