β
β
β
β
β β
β β
π Join our WeChat and Discord
This repo contains PyTorch model definitions, pre-trained weights and inference/sampling code for our paper exploring HunyuanVideo. You can find more visualizations on our project page.
HunyuanVideo: A Systematic Framework For Large Video Generation Model Training
demo.mp4
The video is heavily compressed due to compliance of GitHub policy. The high quality version can be downloaded from here.
- Dec 3, 2024: π€ We release the inference code and model weights of HunyuanVideo.
- HunyuanVideo (Text-to-Video Model)
- Inference
- Checkpoints
- Multi-gpu inference
- Penguin Video Benchmark
- Web Demo (Gradio)
- ComfyUI
- Diffusers
- HunyuanVideo (Image-to-Video Model)
- Inference
- Checkpoints
- HunyuanVideo: A Systematic Framework For Large Video Generation Model Training
We present HunyuanVideo, a novel open-source video foundation model that exhibits performance in video generation that is comparable to, if not superior to, leading closed-source models. In order to train HunyuanVideo model, we adopt several key technologies for model learning, including data curation, image-video joint model training, and an efficient infrastructure designed to facilitate large-scale model training and inference. Additionally, through an effective strategy for scaling model architecture and dataset, we successfully trained a video generative model with over 13 billion parameters, making it the largest among all open-source models.
We conducted extensive experiments and implemented a series of targeted designs to ensure high visual quality, motion diversity, text-video alignment, and generation stability. According to professional human evaluation results, HunyuanVideo outperforms previous state-of-the-art models, including Runway Gen-3, Luma 1.6, and 3 top performing Chinese video generative models. By releasing the code and weights of the foundation model and its applications, we aim to bridge the gap between closed-source and open-source video foundation models. This initiative will empower everyone in the community to experiment with their ideas, fostering a more dynamic and vibrant video generation ecosystem.
HunyuanVideo is trained on a spatial-temporally compressed latent space, which is compressed through Causal 3D VAE. Text prompts are encoded using a large language model, and used as the condition. Gaussian noise and condition are taken as input, our generate model generates an output latent, which is decoded to images or videos through the 3D VAE decoder.
HunyuanVideo introduces the Transformer design and employs a Full Attention mechanism for unified image and video generation. Specifically, we use a "Dual-stream to Single-stream" hybrid model design for video generation. In the dual-stream phase, video and text tokens are processed independently through multiple Transformer blocks, enabling each modality to learn its own appropriate modulation mechanisms without interference. In the single-stream phase, we concatenate the video and text tokens and feed them into subsequent Transformer blocks for effective multimodal information fusion. This design captures complex interactions between visual and semantic information, enhancing overall model performance.
Some previous text-to-video models typically use pretrained CLIP and T5-XXL as text encoders where CLIP uses Transformer Encoder and T5 uses a Encoder-Decoder structure. In constrast, we utilize a pretrained Multimodal Large Language Model (MLLM) with a Decoder-Only structure as our text encoder, which has following advantages: (i) Compared with T5, MLLM after visual instruction finetuning has better image-text alignment in the feature space, which alleviates the difficulty of instruction following in diffusion models; (ii) Compared with CLIP, MLLM has been demonstrated superior ability in image detail description and complex reasoning; (iii) MLLM can play as a zero-shot learner by following system instructions prepended to user prompts, helping text features pay more attention to key information. In addition, MLLM is based on causal attention while T5-XXL utilizes bidirectional attention that produces better text guidance for diffusion models. Therefore, we introduce an extra bidirectional token refiner for enhancing text features.
HunyuanVideo trains a 3D VAE with CausalConv3D to compress pixel-space videos and images into a compact latent space. We set the compression ratios of video length, space and channel to 4, 8 and 16 respectively. This can significantly reduce the number of tokens for the subsequent diffusion transformer model, allowing us to train videos at the original resolution and frame rate.
To address the variability in linguistic style and length of user-provided prompts, we fine-tune the Hunyuan-Large model as our prompt rewrite model to adapt the original user prompt to model-preferred prompt.
We provide two rewrite modes: Normal mode and Master mode, which can be called using different prompts. The prompts are shown here. The Normal mode is designed to enhance the video generation model's comprehension of user intent, facilitating a more accurate interpretation of the instructions provided. The Master mode enhances the description of aspects such as composition, lighting, and camera movement, which leans towards generating videos with a higher visual quality. However, this emphasis may occasionally result in the loss of some semantic details.
The Prompt Rewrite Model can be directly deployed and inferred using the Hunyuan-Large original code. We release the weights of the Prompt Rewrite Model here.
To evaluate the performance of HunyuanVideo, we selected five strong baselines from closed-source video generation models. In total, we utilized 1,533 text prompts, generating an equal number of video samples with HunyuanVideo in a single run. For a fair comparison, we conducted inference only once, avoiding any cherry-picking of results. When comparing with the baseline methods, we maintained the default settings for all selected models, ensuring consistent video resolution. Videos were assessed based on three criteria: Text Alignment, Motion Quality, and Visual Quality. More than 60 professional evaluators performed the evaluation. Notably, HunyuanVideo demonstrated the best overall performance, particularly excelling in motion quality. Please note that the evaluation is based on Hunyuan Video's high-quality version. This is different from the currently released fast version.
| Model | Open Source | Duration | Text Alignment | Motion Quality | Visual Quality | Overall | Ranking |
|---|---|---|---|---|---|---|---|
| HunyuanVideo (Ours) | β | 5s | 61.8% | 66.5% | 95.7% | 41.3% | 1 |
| CNTopA (API) | β | 5s | 62.6% | 61.7% | 95.6% | 37.7% | 2 |
| CNTopB (Web) | β | 5s | 60.1% | 62.9% | 97.7% | 37.5% | 3 |
| GEN-3 alpha (Web) | β | 6s | 47.7% | 54.7% | 97.5% | 27.4% | 4 |
| Luma1.6 (API) | β | 5s | 57.6% | 44.2% | 94.1% | 24.8% | 6 |
| CNTopC (Web) | β | 5s | 48.4% | 47.2% | 96.3% | 24.6% | 5 |
The following table shows the requirements for running HunyuanVideo model (batch size = 1) to generate videos:
| Model | Setting (height/width/frame) |
Denoising step | GPU Peak Memory |
|---|---|---|---|
| HunyuanVideo | 720px1280px129f | 30 | 60GB |
| HunyuanVideo | 544px960px129f | 30 | 45GB |
- An NVIDIA GPU with CUDA support is required.
- We have tested on a single H800/H20 GPU.
- Minimum: The minimum GPU memory required is 60GB for 720px1280px129f and 45G for 544px960px129f.
- Recommended: We recommend using a GPU with 80GB of memory for better generation quality.
- Tested operating system: Linux
Begin by cloning the repository:
git clone https://github.com/tencent/HunyuanVideo
cd HunyuanVideoWe provide an environment.yml file for setting up a Conda environment.
Conda's installation instructions are available here.
We recommend CUDA versions 11.8 and 12.0+.
# 1. Prepare conda environment
conda env create -f environment.yml
# 2. Activate the environment
conda activate HunyuanVideo
# 3. Install pip dependencies
python -m pip install -r requirements.txt
# 4. Install flash attention v2 for acceleration (requires CUDA 11.8 or above)
python -m pip install ninja
python -m pip install git+https://github.com/Dao-AILab/flash-attention.git@v2.5.9.post1Additionally, HunyuanVideo also provides a pre-built Docker image: docker_hunyuanvideo.
# 1. Use the following link to download the docker image tar file (For CUDA 12).
wget https://aivideo.hunyuan.tencent.com/download/HunyuanVideo/hunyuan_video_cu12.tar
# 2. Import the docker tar file and show the image meta information (For CUDA 12).
docker load -i hunyuan_video.tar
docker image ls
# 3. Run the container based on the image
docker run -itd --gpus all --init --net=host --uts=host --ipc=host --name hunyuanvideo --security-opt=seccomp=unconfined --ulimit=stack=67108864 --ulimit=memlock=-1 --privileged docker_image_tagThe details of download pretrained models are shown here.
We list the height/width/frame settings we support in the following table.
| Resolution | h/w=9:16 | h/w=16:9 | h/w=4:3 | h/w=3:4 | h/w=1:1 |
|---|---|---|---|---|---|
| 540p | 544px960px129f | 960px544px129f | 624px832px129f | 832px624px129f | 720px720px129f |
| 720p (recommended) | 720px1280px129f | 1280px720px129f | 1104px832px129f | 832px1104px129f | 960px960px129f |
cd HunyuanVideo
python3 sample_video.py \
--video-size 720 1280 \
--video-length 129 \
--infer-steps 50 \
--prompt "A cat walks on the grass, realistic style." \
--flow-reverse \
--use-cpu-offload \
--save-path ./resultsWe list some more useful configurations for easy usage:
| Argument | Default | Description |
|---|---|---|
--prompt |
None | The text prompt for video generation |
--video-size |
720 1280 | The size of the generated video |
--video-length |
129 | The length of the generated video |
--infer-steps |
50 | The number of steps for sampling |
--embedded-cfg-scale |
6.0 | Embeded Classifier free guidance scale |
--flow-shift |
7.0 | Shift factor for flow matching schedulers |
--flow-reverse |
False | If reverse, learning/sampling from t=1 -> t=0 |
--seed |
None | The random seed for generating video, if None, we init a random seed |
--use-cpu-offload |
False | Use CPU offload for the model load to save more memory, necessary for high-res video generation |
--save-path |
./results | Path to save the generated video |
If you find HunyuanVideo useful for your research and applications, please cite using this BibTeX:
@misc{kong2024hunyuanvideo,
title={HunyuanVideo: A Systematic Framework For Large Video Generative Models},
author={Weijie Kong, Qi Tian, Zijian Zhang, Rox Min, Zuozhuo Dai, Jin Zhou, Jiangfeng Xiong, Xin Li, Bo Wu, Jianwei Zhang, Kathrina Wu, Qin Lin, Aladdin Wang, Andong Wang, Changlin Li, Duojun Huang, Fang Yang, Hao Tan, Hongmei Wang, Jacob Song, Jiawang Bai, Jianbing Wu, Jinbao Xue, Joey Wang, Junkun Yuan, Kai Wang, Mengyang Liu, Pengyu Li, Shuai Li, Weiyan Wang, Wenqing Yu, Xinchi Deng, Yang Li, Yanxin Long, Yi Chen, Yutao Cui, Yuanbo Peng, Zhentao Yu, Zhiyu He, Zhiyong Xu, Zixiang Zhou, Zunnan Xu, Yangyu Tao, Qinglin Lu, Songtao Liu, Dax Zhou, Hongfa Wang, Yong Yang, Di Wang, Yuhong Liu, and Jie Jiang, along with Caesar Zhong},
year={2024},
archivePrefix={arXiv},
primaryClass={cs.CV}
}We would like to thank the contributors to the SD3, FLUX, Llama, LLaVA, Xtuner, diffusers and huggingface repositories, for their open research and exploration. Additionally, we also thank the Tencent Hunyuan Multimodal team for their help with the text encoder.
