Hunyuan Video gguf

Introduction

HunyuanVideo-GGUF is a text-to-video model derived from the tencent/HunyuanVideo model. It uses the GGUF library and is quantized by city96. The model is designed to integrate with ComfyUI, specifically through the HunyuanVideo nodes.

Architecture

The model is a direct GGUF conversion of the tencent/HunyuanVideo model. It operates with the native ComfyUI nodes and is not a fine-tuned variant, meaning it retains the original license and restrictions.

Training

This model is quantized, which means it has been processed to use reduced precision without fine-tuning, allowing for efficient computation while retaining performance. Refer to the provided chart for understanding quantization types and their implications.

Guide: Running Locally

1. Installation:

   • Clone the ComfyUI-GGUF repository from GitHub.
   • Download the HunyuanVideo model files and place them in the ComfyUI/models/unet directory.
   • Download the VAE from Kijai's repository.

2. Execution:

   • Ensure you have all dependencies installed as specified in the ComfyUI-GGUF GitHub readme.
   • Run the model using the ComfyUI interface.

3. Cloud GPUs:

   • For optimal performance, consider using cloud GPU services such as AWS, Google Cloud, or Azure to handle the computational load.

License

The model is distributed under the tencent-hunyuan-community license, which imposes certain restrictions. It is important to review the LICENSE.md file for specific terms and conditions.