Enhances user input for Stable Diffusion v1-4

The online demo at Hugging Face Spaces uses CPU, so the generation speed will be slow. Please load the model locally with GPUs for faster generation.