|
|
--- |
|
|
library_name: pytorch |
|
|
license: mit |
|
|
pipeline_tag: image-to-video |
|
|
tags: |
|
|
- portrait-animation |
|
|
- video-generation |
|
|
- keypoint-based |
|
|
- efficient |
|
|
--- |
|
|
|
|
|
<h1 align="center">LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control</h1> |
|
|
|
|
|
<div align='center'> |
|
|
<a href='https://github.com/cleardusk' target='_blank'><strong>Jianzhu Guo</strong></a><sup> 1*β </sup>  |
|
|
<a href='https://github.com/Mystery099' target='_blank'><strong>Dingyun Zhang</strong></a><sup> 1,2*</sup>  |
|
|
<a href='https://github.com/KwaiVGI' target='_blank'><strong>Xiaoqiang Liu</strong></a><sup> 1</sup>  |
|
|
<a href='https://github.com/zzzweakman' target='_blank'><strong>Zhizhou Zhong</strong></a><sup> 1,3</sup>  |
|
|
<a href='https://scholar.google.com.hk/citations?user=_8k1ubAAAAAJ' target='_blank'><strong>Yuan Zhang</strong></a><sup> 1</sup>  |
|
|
</div> |
|
|
|
|
|
<div align='center'> |
|
|
<a href='https://scholar.google.com/citations?user=P6MraaYAAAAJ' target='_blank'><strong>Pengfei Wan</strong></a><sup> 1</sup>  |
|
|
<a href='https://openreview.net/profile?id=~Di_ZHANG3\' target='_blank'><strong>Di Zhang</strong></a><sup> 1</sup>  |
|
|
</div> |
|
|
|
|
|
<div align='center'> |
|
|
<sup>1 </sup>Kuaishou Technology  <sup>2 </sup>University of Science and Technology of China  <sup>3 </sup>Fudan University  |
|
|
</div> |
|
|
<div align='center'> |
|
|
<small><sup>*</sup> Equal contributions</small> |
|
|
<small><sup>β </sup> Corresponding author</small> |
|
|
</div> |
|
|
|
|
|
<div align="center" style="display: flex; justify-content: center; flex-wrap: wrap;"> |
|
|
<!-- <a href='LICENSE'><img src='https://img.shields.io/badge/license-MIT-yellow'></a> --> |
|
|
<a href='https://arxiv.org/pdf/2407.03168'><img src='https://img.shields.io/badge/arXiv-LivePortrait-red'></a> |
|
|
<a href='https://liveportrait.github.io'><img src='https://img.shields.io/badge/Project-LivePortrait-green'></a> |
|
|
<a href='https://huggingface.co/spaces/KwaiVGI/liveportrait'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue'></a> |
|
|
<a href="https://github.com/KwaiVGI/LivePortrait"><img src="https://img.shields.io/github/stars/KwaiVGI/LivePortrait"></a> |
|
|
</div> |
|
|
<br> |
|
|
|
|
|
<p align="center"> |
|
|
<img src="./docs/showcase2.gif" alt="showcase"> |
|
|
π₯ For more results, visit our <a href="https://liveportrait.github.io/"><strong>homepage</strong></a> π₯ |
|
|
</p> |
|
|
|
|
|
|
|
|
|
|
|
## π₯ Updates |
|
|
- **`2025/01/01`**: πΆ We updated a new version of the Animals model with more data, see [**here**](./assets/docs/changelog/2025-01-01.md). |
|
|
- **`2024/10/18`**: β We have updated the versions of the `transformers` and `gradio` libraries to avoid security vulnerabilities. Details [here](https://github.com/KwaiVGI/LivePortrait/pull/421/files). |
|
|
- **`2024/08/29`**: π¦ We update the Windows [one-click installer](https://huggingface.co/cleardusk/LivePortrait-Windows/blob/main/LivePortrait-Windows-v20240829.zip) and support auto-updates, see [changelog](https://huggingface.co/cleardusk/LivePortrait-Windows#20240829). |
|
|
- **`2024/08/19`**: πΌοΈ We support **image driven mode** and **regional control**. For details, see [**here**](./assets/docs/changelog/2024-08-19.md). |
|
|
- **`2024/08/06`**: π¨ We support **precise portrait editing** in the Gradio interface, inspired by [ComfyUI-AdvancedLivePortrait](https://github.com/PowerHouseMan/ComfyUI-AdvancedLivePortrait). See [**here**](./assets/docs/changelog/2024-08-06.md). |
|
|
- **`2024/08/05`**: π¦ Windows users can now download the [one-click installer](https://huggingface.co/cleardusk/LivePortrait-Windows/blob/main/LivePortrait-Windows-v20240806.zip) for Humans mode and **Animals mode** now! For details, see [**here**](./assets/docs/changelog/2024-08-05.md). |
|
|
- **`2024/08/02`**: πΈ We released a version of the **Animals model**, along with several other updates and improvements. Check out the details [**here**](./assets/docs/changelog/2024-08-02.md)! |
|
|
- **`2024/07/25`**: π¦ Windows users can now download the package from [HuggingFace](https://huggingface.co/cleardusk/LivePortrait-Windows/tree/main) or [BaiduYun](https://pan.baidu.com/s/1FWsWqKe0eNfXrwjEhhCqlw?pwd=86q2). Simply unzip and double-click `run_windows.bat` to enjoy!\n- **`2024/07/24`**: π¨ We support pose editing for source portraits in the Gradio interface. Weβve also lowered the default detection threshold to increase recall. [Have fun](https://github.com/KwaiVGI/LivePortrait/blob/main/assets/docs/changelog/2024-07-24.md)!\n- **`2024/07/19`**: β¨ We support ποΈ portrait video editing (aka v2v)! More to see [here](https://github.com/KwaiVGI/LivePortrait/blob/main/assets/docs/changelog/2024-07-19.md).\n- **`2024/07/17`**: π We support macOS with Apple Silicon, modified from [jeethu](https://github.com/jeethu)\'s PR [#143](https://github.com/KwaiVGI/LivePortrait/pull/143).\n- **`2024/07/10`**: πͺ We support audio and video concatenating, driving video auto-cropping, and template making to protect privacy. More to see [here](https://github.com/KwaiVGI/LivePortrait/blob/main/assets/docs/changelog/2024-07-10.md).\n- **`2024/07/09`**: π€ We released the [HuggingFace Space](https://huggingface.co/spaces/KwaiVGI/liveportrait), thanks to the HF team and [Gradio](https://github.com/gradio-app/gradio)!\n- **`2024/07/04`**: π We released the initial version of the inference code and models. Continuous updates, stay tuned!\n- **`2024/07/04`**: π₯ We released the [homepage](https://liveportrait.github.io) and technical report on [arXiv](https://arxiv.org/pdf/2407.03168).\n\n\n## Introduction π\nThis repo, named **LivePortrait**, contains the official PyTorch implementation of our paper [LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control](https://arxiv.org/pdf/2407.03168).\nWe are actively updating and improving this repository. If you find any bugs or have suggestions, welcome to raise issues or submit pull requests (PR) π.\n\n## Getting Started π\n### 1. Clone the code and prepare the environment\n```bash\ngit clone https://github.com/KwaiVGI/LivePortrait\ncd LivePortrait\n\n# create env using conda\nconda create -n LivePortrait python==3.9\nconda activate LivePortrait\n\n# install dependencies with pip\n# for Linux and Windows users\npip install -r requirements.txt\n# for macOS with Apple Silicon users\npip install -r requirements_macOS.txt\n```\n\n**Note:** make sure your system has [FFmpeg](https://ffmpeg.org/download.html) installed, including both `ffmpeg` and `ffprobe`!\n\n### 2. Download pretrained weights\n\nThe easiest way to download the pretrained weights is from HuggingFace:\n```bash\n# first, ensure git-lfs is installed, see: https://docs.github.com/en/repositories/working-with-files/managing-large-files/installing-git-large-file-storage\ngit lfs install\n# clone and move the weights\ngit clone https://huggingface.co/KwaiVGI/LivePortrait temp_pretrained_weights\nmv temp_pretrained_weights/* pretrained_weights/\nrm -rf temp_pretrained_weights\n```\n\nAlternatively, you can download all pretrained weights from [Google Drive](https://drive.google.com/drive/folders/1UtKgzKjFAOmZkhNK-OYT0caJ_w2XAnib) or [Baidu Yun](https://pan.baidu.com/s/1MGctWmNla_vZxDbEp2Dtzw?pwd=z5cn). Unzip and place them in `./pretrained_weights`.\n\nEnsuring the directory structure is as follows, or contains:\n```text\npretrained_weights\nβββ insightface\nβ βββ models\nβ βββ buffalo_l\nβ βββ 2d106det.onnx\nβ βββ det_10g.onnx\nβββ liveportrait\n βββ base_models\n β βββ appearance_feature_extractor.pth\n β βββ motion_extractor.pth\n β βββ spade_generator.pth\n β βββ warping_module.pth\n βββ landmark.onnx\n βββ retargeting_models\n βββ stitching_retargeting_module.pth\n```\n\n### 3. Inference π\n\n#### Fast hands-on\n```bash\n# For Linux and Windows\npython inference.py\n\n# For macOS with Apple Silicon, Intel not supported, this maybe 20x slower than RTX 4090\nPYTORCH_ENABLE_MPS_FALLBACK=1 python inference.py\n```\n\nIf the script runs successfully, you will get an output mp4 file named `animations/s6--d0_concat.mp4`. This file includes the following results: driving video, input image or video, and generated result.\n\n<p align="center">\n <img src="./docs/inference.gif" alt="image">\n</p>\n\nOr, you can change the input by specifying the `-s` and `-d` arguments:\n\n```bash\n# source input is an image\npython inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d0.mp4\n\n# source input is a video β¨\npython inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d0.mp4\n\n# more options to see\npython inference.py -h\n```\n\n#### Driving video auto-cropping π’π’π’\nTo use your own driving video, we **recommend**: β¬οΈ\n - Crop it to a **1:1** aspect ratio (e.g., 512x512 or 256x256 pixels), or enable auto-cropping by `--flag_crop_driving_video`.\n - Focus on the head area, similar to the example videos.\n - Minimize shoulder movement.\n - Make sure the first frame of driving video is a frontal face with **neutral expression**.\n\nBelow is a auto-cropping case by `--flag_crop_driving_video`:\n```bash\npython inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d13.mp4 --flag_crop_driving_video\n```\n\nIf you find the results of auto-cropping is not well, you can modify the `--scale_crop_driving_video`, `--vy_ratio_crop_driving_video` options to adjust the scale and offset, or do it manually.\n\n#### Motion template making\nYou can also use the auto-generated motion template files ending with `.pkl` to speed up inference, and **protect privacy**, such as:\n```bash\npython inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d5.pkl # portrait animation\npython inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d5.pkl # portrait video editing\n```\n\n### 4. Gradio interface π€\n\nWe also provide a Gradio <a href=\'https://github.com/gradio-app/gradio\'><img src=\'https://img.shields.io/github/stars/gradio-app/gradio\'></a> interface for a better experience, just run by:\n\n```bash\n# For Linux and Windows users (and macOS with Intel??)\npython app.py\n\n# For macOS with Apple Silicon users, Intel not supported, this maybe 20x slower than RTX 4090\nPYTORCH_ENABLE_MPS_FALLBACK=1 python app.py\n```\n\nYou can specify the `--server_port`, `--share`, `--server_name` arguments to satisfy your needs!\n\nπ We also provide an acceleration option `--flag_do_torch_compile`. The first-time inference triggers an optimization process (about one minute), making subsequent inferences 20-30% faster. Performance gains may vary with different CUDA versions.\n```bash\n# enable torch.compile for faster inference\npython app.py --flag_do_torch_compile\n```\n**Note**: This method is not supported on Windows and macOS.\n\n**Or, try it out effortlessly on [HuggingFace](https://huggingface.co/spaces/KwaiVGI/liveportrait) π€**\n\n### 5. Inference speed evaluation πππ\nWe have also provided a script to evaluate the inference speed of each module:\n\n```bash\n# For NVIDIA GPU\npython speed.py\n```\n\nBelow are the results of inferring one frame on an RTX 4090 GPU using the native PyTorch framework with `torch.compile`:\n\n| Model | Parameters(M) | Model Size(MB) | Inference(ms) |\n|-----------------------------------|:-------------:|:--------------:|:-------------:|\n| Appearance Feature Extractor | 0.84 | 3.3 | 0.82 |\n| Motion Extractor | 28.12 | 108 | 0.84 |\n| Spade Generator | 55.37 | 212 | 7.59 |\n| Warping Module | 45.53 | 174 | 5.21 |\n| Stitching and Retargeting Modules | 0.23 | 2.3 | 0.31 |\n\n*Note: The values for the Stitching and Retargeting Modules represent the combined parameter counts and total inference time of three sequential MLP networks.*\n\n## Community Resources π€\n\nDiscover the invaluable resources contributed by our community to enhance your LivePortrait experience:\n\n- [ComfyUI-LivePortraitKJ](https://github.com/kijai/ComfyUI-LivePortraitKJ) by [@kijai](https://github.com/kijai)\n- [comfyui-liveportrait](https://github.com/shadowcz007/comfyui-liveportrait) by [@shadowcz007](https://github.com/shadowcz007)\n- [LivePortrait In ComfyUI](https://www.youtube.com/watch?v=aFcS31OWMjE) by [@Benji](https://www.youtube.com/@TheFutureThinker)\n- [LivePortrait hands-on tutorial](https://www.youtube.com/watch?v=uyjSTAOY7yI) by [@AI Search](https://www.youtube.com/@theAIsearch)\n- [ComfyUI tutorial](https://www.youtube.com/watch?v=8-IcDDmiUMM) by [@Sebastian Kamph](https://www.youtube.com/@sebastiankamph)\n- [Replicate Playground](https://replicate.com/fofr/live-portrait) and [cog-comfyui](https://github.com/fofr/cog-comfyui) by [@fofr](https://github.com/fofr)\n\nAnd many more amazing contributions from our community!\n\n## Acknowledgements π\nWe would like to thank the contributors of [FOMM](https://github.com/AliaksandrSiarohin/first-order-model), [Open Facevid2vid](https://github.com/zhanglonghao1992/One-Shot_Free-View_Neural_Talking_Head_Synthesis), [SPADE](https://github.com/NVlabs/SPADE), [InsightFace](https://github.com/deepinsight/insightface) and [X-Pose](https://github.com/IDEA-Research/X-Pose) repositories, for their open research and contributions.\n\n## Ethics Considerations π‘οΈ\nPortrait animation technologies come with social risks, particularly the potential for misuse in creating deepfakes. To mitigate these risks, itβs crucial to follow ethical guidelines and adopt responsible usage practices. At present, the synthesized results contain visual artifacts that may help in detecting deepfakes. Please note that we do not assume any legal responsibility for the use of the results generated by this project.\n\n## Citation π\nIf you find LivePortrait useful for your research, welcome to π this repo and cite our work using the following BibTeX:\n```bibtex\n@article{guo2024liveportrait,\n title = {LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control},\n author = {Guo, Jianzhu and Zhang, Dingyun and Liu,Xiaoqiang and Zhong, Zhizhou and Zhang, Yuan and Wan, Pengfei and Zhang, Di},\n journal = {arXiv preprint arXiv:2407.03168},\n year = {2024}\n}\n```\n\n*Long live in arXiv.*\n\n## Contact π§\n[**Jianzhu Guo (ιε»Ίη )**](https://guojianzhu.com); **guojianzhu1994@gmail.com** |
