Hunyuan Video 6 Steps ComfyUI Workflow Low VRAM

With Hunyuan text-to-video technology, you can now generate videos in just 6 simple steps. Yes, you read that right—only six!

Here’s how you can create your own Hunyuan video in just a few steps:

1. Download the Workflow
   Visit ComfyUI Blog and download the pre-built workflow tailored for Hunyuan video creation.
2. Load ComfyUI
   Launch ComfyUI on your system and load the downloaded workflow. For reference, this tutorial uses a system with 12GB of VRAM.
3. Gather Required Files
   You’ll need the following files:
   • pytorch_model.pt (Save this in the VAE folder of ComfyUI.)
   • mp_rank_00_model_states.pt (Place this in the Diffusion Model folder.)
   • Clip I of Flux and Llava Llama3 FP8 (Save these in the Text Encoder folder.)
4. Load the Files into ComfyUI
   • In the VAE loader, select pytorch_model.pt.
   • In the Diffusion Model loader, select mp_rank_00_model_states.pt. If you encounter memory issues, switch the Weight Dtype to FP8.
   • For Dual Clip Load, select Clip I of Flux and Llava Llama3 FP8, and set the type to Hunyuan Video.
5. Configure the Workflow
   • Add your text prompt in the Clip Text Encoder and set Flux Guidance to 10 and Model Sampling to 7.
   • In the Empty Latent section, set the resolution to 848×480 for standard output or 720×1224 for HD.
   • In the Basic Scheduler, select the Simple Scheduler and set the steps to 6 for quick results or up to 20 for enhanced quality.

• VAE: pytorch_model.pt
• Diffusion Model: mp_rank_00_model_states.pt
• Text Encoder: llava_llama3_fp8_scaled.safetensors
• Clip I: clip_l.safetensors