LoRA trained using Marie Antoinette stills as a training data set. It is overturned so it will produce results even at low values such as 0.5.