This is a Lora model trained based on stable diffusion v1.5. The dataset is from a comic book. I guarantee that there is no problem with the dataset. Maybe there is a problem with the generated description. I will see how to improve it next time.