Re-Train of the previous model with added upscaled photos and generated with high res fix from ControlNet so the results should be more clear and versatile.