numz/ComfyUI-SeedVR2_VideoUpscaler ? reverse-engineered prompt

Build me a ComfyUI extension that can take a low quality video or image and upscale it into a cleaner, sharper result with good detail and fewer artifacts. I want it to work inside ComfyUI with an easy workflow, and also have a standalone command line mode for batch processing videos if that is supported. It should handle long videos without running out of memory, and it would be great if it can use multiple GPUs when available.

Please make it practical to install, include sensible example workflows, and keep the setup clear for people who just want to drop it into ComfyUI and start using it. If there are any model files or special requirements, make sure the project reads nicely and the defaults are set up so I can get a result without a lot of tweaking. If needed, look up the current docs online so the implementation matches the latest release behavior.