Hunyuan Video

Hunyuan Video is a new, state of the art, AI Video Generator that creates high-quality videos from text descriptions. With 13B parameters and state-of-the-art performance, it's the most powerful open-source video generation model available.

Get Started Learn More

About Hunyuan Video

HunyuanVideo is a groundbreaking open-source video generation model built by Tencent. This project represents a significant advancement in AI-powered video creation, offering performance comparable to leading closed-source models.

With HunyuanVideo, every video is generated from text descriptions, powered by over 13 billion parameters, making it the largest open-source video generation model available today. The model excels in producing high-quality videos with outstanding motion diversity and text-video alignment.

Thanks to our innovative architecture and advanced technologies like MLLM Text Encoder and 3D VAE, HunyuanVideo achieves remarkable video generation capabilities. The model supports various resolutions up to 720p×1280p, delivering exceptional visual quality and natural motion.

Optimized for modern GPUs, HunyuanVideo represents a major step forward in democratizing video generation technology, making professional-grade video creation accessible to everyone through open-source software.

Command Example

Links

Read Blog

Learn about Hunyuan Video AI

View Code

Explore the open-source codebase on GitHub

Download Weights

Get the model weights from Hugging Face

Documentation

Access detailed technical documentation

Join Our Community

💬

Discord

Join our Discord server to share creations, get help, and connect with other players.

Join Discord
🐙

GitHub

Contribute to the project, report issues, or explore the code on GitHub.

View Repository

Frequently Asked Questions

What are the system requirements? +
The minimum GPU memory required is 60GB for 720p×1280p videos and 45GB for 544p×960p videos. An NVIDIA GPU with CUDA support is required, and we recommend using a GPU with 80GB of memory for optimal performance.
How do I get started with HunyuanVideo? +
You can get started by downloading the code and model weights from our GitHub repository. We provide comprehensive documentation and installation guides for both Linux and Docker environments.
What types of videos can it generate? +
HunyuanVideo can generate a wide range of videos from text descriptions, including natural scenes, human actions, animations, and more. It supports various resolutions and aspect ratios to suit different needs.
Is it really open source? +
Yes! HunyuanVideo is completely open source. The code, model weights, and documentation are freely available on GitHub and Hugging Face. You can use, modify, and distribute it under the open-source license.
How does it compare to other video generation models? +
According to professional evaluations, HunyuanVideo outperforms previous state-of-the-art models in text alignment (68.5%), motion quality (64.5%), and visual quality (96.4%), making it the leading open-source video generation model.
Can I integrate it into my own applications? +
Yes, HunyuanVideo can be integrated into your applications. We provide comprehensive API documentation and examples for both inference and fine-tuning to help you get started.
What makes HunyuanVideo unique? +
HunyuanVideo stands out for its 13B parameters, advanced MLLM text encoder, 3D VAE architecture, and superior performance metrics. It's the largest and most capable open-source video generation model available.