The LoRA has been trained on a dataset made of photos therefore, it tend to produce photo-realistic pictures. It can however produce good painting or illustration but it depends on the painting/illustration style LoRA, some work better than others here.