Official <a target="_blank" rel="ugc" href="https://docs.comfy.org/installation/comfyui_portable_windows#download-comfyui-portable">ComfyUI</a> workflows with fl2va int8 version of the <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main">Minimax H3</a> model generating 8 sec 720p video on rtx4080 16GB + shared RAM: 32GBwith no optimizations generates for 25 min<hr />with <a target="_blank" rel="ugc" href="https://github.com/duckyshell/ComfyUI-MiniMaxH3-FirstBlockCache">FirstBlockCache</a> reduces generation to 15 min - lossless quality<ul><li>just works or set mode to H3 Aggressive for even more speed</li></ul><hr />with <a target="_blank" rel="ugc" href="https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo">Turbo Lora</a> reduces generation to 6 min - with latest version the quality loss is minimal on 6 steps<hr />Other changes from official workflow:<ul><li>replaced Load Image node with Image Crop for convenience to make the image in the video aspect ratio to prevent stretching</li><li>replaced Save Video with VHS version of it to improve the codec quality of the video file</li><li>added notification node to alert when generation is finished and display generation time</li><li>added Comfy Kitchen Attention node for optimization</li></ul>Updates:<ul><li>changed all models to int8 - better quality</li><li>updated turbo lora - better quality and speed</li><li>added RIFE VFI frame interpolation and <a rel="ugc" href="https://github.com/whmc76/ComfyUI-NVIDIA-RTX-VSR-Pro">RTX VSR Pro</a> upscale</li></ul>

Minimax H3 i2v Turbo / FirstBlockCache

After tests updates:<ul><li>best model versions</li><li>best settings</li></ul>based on minmax speed and quality, trying to avoid limits on rtx 4080 16GB VRAM + shared 32GB RAM

Files with SHA256: 7bf167554a954969554bf0dd7a51d44b8aa009aa180f8d7fac422983466715aa