• Download from https://huggingface.co/stabilityai/stable-diffusion-xl-base-0.9 . You can fill out the form with anything.
• Download ComfyUI from https://github.com/comfyanonymous/ComfyUI .
• Unzip ComfyUI.
• Copy the downloaded sd_xl_base_0.9.safetensors to \ComfyUI_windows_portable\ComfyUI\models\checkpoints .
• Run run_nvidia_gpu.bat .
• Change both width and height under Empty Latent Image to 1024.
• At Ksampler Change the sampler name to dpmpp_2s_ancestral
• Enter positive prompt
• Enter negative prompt (optional)

Adding the refiner to stable-diffusion-xl-base-0.9:

• Download sd_xl_refiner_0.9.safetensors from https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-0.9/tree/main
• copy sd_xl_refiner_0.9.safetensors to \ComfyUI_windows_portable\ComfyUI\models\checkpoints .
• Download https://github.com/camenduru/sdxl-colab/blob/main/refiner.json
• Load refiner.json with ComfyUI