WaveSpeed is an all-in-one inference optimization solution for ComfyUI. It implements dynamic caching (First Block Cache) and enhanced torch.compile to accelerate diffusion model inference. The tool supports multiple model families including FLUX.1-dev, Wan 2.1 (text-to-video and image-to-video), LTXV, HunyuanVideo, and SD3.5/SDXL. It provides custom nodes and demo workflows for easy integration.