Note to users: Using the likeness or personal information of a real person in Low-Rank Adaptation (LoRA) applications comes with a responsibility to uphold ethical standards. Respect their privacy, rights, and dignity. Any misrepresentation, manipulation, or any form of harm that may arise from the use of this LoRA is YOUR responsibility and not of the creator of this LoRA.

This is still experimental and PDXL is far more moody than SD1.5 or SDXL - so YMMV.

Actually works quite well without a trigger word as well, followed by GFPGAN (strength:0.25) Upscale to remove artifacts. Check second set of posted images.

Based on a suggestion made by @Malessar , this has been trained on the original Pony model and so should therefore transfer to other Pony based models a little more broadly.