This ComfyUI custom node adds Virtual VRAM management and MultiGPU integration for image and video generation models. It uses DisTorch to distribute model components (UNet, CLIP, VAE) across multiple GPUs or between VRAM and DRAM to maximize available memory for latent space processing. Includes bespoke nodes for WanVideoWrapper and supports both .safetensors and GGUF model formats with model-driven allocation options.