Probably ComfyUI is one of the easiest way to get started with local image/video models. Or perhaps vLLM, if they have support for it already, would be something like `vllm serve <model> --omni --port 9080`