A T2V workflow for FastHunyuanVideo.

This workflow is based on HunyuanVideoWrapper, as the standard workflow in ComfyUI currently generates blurry images when using the Fast model.

The workflow only requires 6 steps. It takes approximately 103 seconds to generate a 720x480 resolution video with 81 frames, using about 16-18GB of VRAM.

Model Requirements

Unet (/ComfyUI/Models/Unet)

https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_FastVideo_720_fp8_e4m3fn.safetensors

Vae  (/ComfyUI/Models/Vae/KVae)

https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors

Note: The Vae in this workflow must be placed in a specific folder. Please adjust the path to match your setup.