mirror of
https://github.com/KwaiVGI/LivePortrait.git
synced 2024-12-22 20:42:38 +00:00
288 lines
14 KiB
Markdown
288 lines
14 KiB
Markdown
<h1 align="center">LivePortrait:通过拼接和重定向控制实现高效的人像动画</h1>
|
||
|
||
<div align='center'>
|
||
<a href='https://github.com/cleardusk' target='_blank'><strong>Jianzhu Guo</strong></a><sup> 1†</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> Corresponding author</small>
|
||
</div>
|
||
|
||
<br>
|
||
<div align="center">
|
||
<!-- <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>
|
||
<br>
|
||
<a href="../../readme.md"><strong>English</strong></a> | <strong>简体中文</strong>
|
||
</div>
|
||
|
||
<br>
|
||
|
||
|
||
<p align="center">
|
||
<img src="./assets/docs/showcase2.gif" alt="showcase">
|
||
<br>
|
||
🔥 更多效果,请查看我们的 <a href="https://liveportrait.github.io/"><strong>主页</strong></a> 🔥
|
||
</p>
|
||
|
||
|
||
|
||
## 🔥 更新日志
|
||
- **`2024/08/06`**:🎨 我们在Gradio界面支持**精确的人像编辑**, 受到[ComfyUI-AdvancedLivePortrait](https://github.com/PowerHouseMan/ComfyUI-AdvancedLivePortrait)启发。详见[**这里**](./assets/docs/changelog/2024-08-06.md)。
|
||
- **`2024/08/05`**:📦Windows用户现在可以下载[一键安装程序](https://huggingface.co/cleardusk/LivePortrait-Windows/blob/main/LivePortrait-Windows-v20240806.zip),支持**人类模式**和**动物模式**!详情见[**这里**](./assets/docs/changelog/2024-08-05.md)。
|
||
- **`2024/08/02`**:😸 我们发布了**动物模型**版本,以及其他一些更新和改进。查看详情[**这里**](./assets/docs/changelog/2024-08-02.md)!
|
||
- **`2024/07/25`**:📦 Windows用户现在可以从 [HuggingFace](https://huggingface.co/cleardusk/LivePortrait-Windows/tree/main) 或 [百度云](https://pan.baidu.com/s/1FWsWqKe0eNfXrwjEhhCqlw?pwd=86q2) 下载软件包。解压并双击`run_windows.bat`即可享受!
|
||
- **`2024/07/24`**:🎨 我们在Gradio界面支持源人像的姿势编辑。我们还降低了默认检测阈值以增加召回率。[玩得开心](assets/docs/changelog/2024-07-24.md)!
|
||
- **`2024/07/19`**:✨ 我们支持🎞️ **人像视频编辑(aka v2v)**!更多信息见[**这里**](assets/docs/changelog/2024-07-19.md)。
|
||
- **`2024/07/17`**:🍎 我们支持macOS搭载Apple Silicon,修改来自 [jeethu](https://github.com/jeethu) 的PR [#143](https://github.com/KwaiVGI/LivePortrait/pull/143) 。
|
||
- **`2024/07/10`**:💪我们支持音频和视频拼接、驱动视频自动裁剪以及制作模板以保护隐私。更多信息见[这里](assets/docs/changelog/2024-07-10.md)。
|
||
- **`2024/07/09`**:🤗 我们发布了[HuggingFace Space](https://huggingface.co/spaces/KwaiVGI/liveportrait),感谢HF团队和[Gradio](https://github.com/gradio-app/gradio)!
|
||
- **`2024/07/04`**:😊 我们发布了初始版本的推理代码和模型。持续更新,敬请关注!
|
||
- **`2024/07/04`**:🔥 我们发布了[主页](https://liveportrait.github.io)和在[arXiv](https://arxiv.org/pdf/2407.03168)上的技术报告。
|
||
|
||
|
||
|
||
## 介绍 📖
|
||
此仓库名为**LivePortrait**,包含我们论文([LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control](https://arxiv.org/pdf/2407.03168))的官方PyTorch实现。 我们正在积极更新和改进此仓库。如果您发现任何错误或有建议,欢迎提出问题或提交合并请求💖。
|
||
|
||
## 上手指南 🏁
|
||
### 1. 克隆代码和安装运行环境 🛠️
|
||
|
||
> [!Note]
|
||
> 确保您的系统已安装[`git`](https://git-scm.com/)、[`conda`](https://anaconda.org/anaconda/conda)和[`FFmpeg`](https://ffmpeg.org/download.html)。有关FFmpeg安装的详细信息,见[**如何安装FFmpeg**](assets/docs/how-to-install-ffmpeg.md)。
|
||
|
||
```bash
|
||
git clone https://github.com/KwaiVGI/LivePortrait
|
||
cd LivePortrait
|
||
|
||
# 使用conda创建环境
|
||
conda create -n LivePortrait python=3.9
|
||
conda activate LivePortrait
|
||
```
|
||
|
||
#### 对于Linux或Windows用户
|
||
|
||
[X-Pose](https://github.com/IDEA-Research/X-Pose)需要您的`torch`版本与CUDA版本兼容。
|
||
|
||
首先,通过以下命令检查您当前的CUDA版本:
|
||
|
||
```bash
|
||
nvcc -V # example versions: 11.1, 11.8, 12.1, etc.
|
||
```
|
||
|
||
然后,安装相应版本的torch。以下是不同CUDA版本的示例。如果您的CUDA版本未列出,请访问[PyTorch官方网站](https://pytorch.org/get-started/previous-versions)获取安装命令:
|
||
```bash
|
||
# for CUDA 11.1
|
||
pip install torch==1.10.1+cu111 torchvision==0.11.2 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
|
||
# for CUDA 11.8
|
||
pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu118
|
||
# for CUDA 12.1
|
||
pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu121
|
||
# ...
|
||
```
|
||
|
||
最后,安装其余依赖项:
|
||
|
||
```bash
|
||
pip install -r requirements.txt
|
||
```
|
||
|
||
#### 对于搭载Apple Silicon的macOS用户
|
||
|
||
[X-Pose](https://github.com/IDEA-Research/X-Pose)依赖项不支持macOS,因此您可以跳过其安装。人类模式照常工作,但不支持动物模式。使用为搭载Apple Silicon的macOS提供的requirements文件:
|
||
|
||
```bash
|
||
# 对于搭载Apple Silicon的macOS用户
|
||
pip install -r requirements_macOS.txt
|
||
```
|
||
|
||
### 2. 下载预训练权重(Pretrained weights) 📥
|
||
|
||
从HuggingFace下载预训练权重的最简单方法是:
|
||
```bash
|
||
# !pip install -U "huggingface_hub[cli]"
|
||
huggingface-cli download KwaiVGI/LivePortrait --local-dir pretrained_weights --exclude "*.git*" "README.md" "docs"
|
||
```
|
||
|
||
若您不能访问嬉皮笑脸平台(Huggingface),你可以访问其镜像网站[hf-mirror](https://hf-mirror.com/)进行下载操作:
|
||
|
||
```bash
|
||
# !pip install -U "huggingface_hub[cli]"
|
||
export HF_ENDPOINT=https://hf-mirror.com
|
||
huggingface-cli download KwaiVGI/LivePortrait --local-dir pretrained_weights --exclude "*.git*" "README.md" "docs"
|
||
```
|
||
|
||
或者,您可以从[Google Drive](https://drive.google.com/drive/folders/1UtKgzKjFAOmZkhNK-OYT0caJ_w2XAnib)或[百度云](https://pan.baidu.com/s/1MGctWmNla_vZxDbEp2Dtzw?pwd=z5cn)(进行中)下载所有预训练权重。解压并将它们放置在`./pretrained_weights`目录下。
|
||
|
||
确保目录结构如所示包含[**本仓库该路径**](assets/docs/directory-structure.md)其中展示的内容。
|
||
|
||
### 3. 推理 🚀
|
||
|
||
#### 快速上手(人类模型)👤
|
||
|
||
```bash
|
||
# 对于Linux和Windows用户
|
||
python inference.py
|
||
|
||
# 对于搭载Apple Silicon的macOS用户(Intel未测试)。注意:这可能比RTX 4090慢20倍
|
||
PYTORCH_ENABLE_MPS_FALLBACK=1 python inference.py
|
||
```
|
||
|
||
如果脚本成功运行,您将得到一个名为`animations/s6--d0_concat.mp4`的输出mp4文件。此文件包含以下结果:驱动视频、输入图像或视频以及生成结果。
|
||
|
||
<p align="center">
|
||
<img src="./assets/docs/inference.gif" alt="image">
|
||
</p>
|
||
或者,您可以通过指定`-s`和`-d`参数来更改输入:
|
||
|
||
```bash
|
||
# 源输入是图像
|
||
python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d0.mp4
|
||
|
||
# 源输入是视频 ✨
|
||
python inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d0.mp4
|
||
|
||
# 更多选项请见
|
||
python inference.py -h
|
||
```
|
||
|
||
#### 快速上手(动物模型) 🐱🐶
|
||
|
||
动物模式仅在Linux和Windows上经过测试,并且需要NVIDIA GPU。
|
||
|
||
您需要首先构建一个名为`MultiScaleDeformableAttention`的OP,该OP由[X-Pose](https://github.com/IDEA-Research/X-Pose)使用,这是一个通用的关键点检测框架。
|
||
|
||
```bash
|
||
cd src/utils/dependencies/XPose/models/UniPose/ops
|
||
python setup.py build install
|
||
cd - # 等同于 cd ../../../../../../../
|
||
```
|
||
|
||
然后执行
|
||
```bash
|
||
python inference_animals.py -s assets/examples/source/s39.jpg -d assets/examples/driving/wink.pkl --driving_multiplier 1.75 --no_flag_stitching
|
||
```
|
||
如果脚本成功运行,您将得到一个名为`animations/s39--wink_concat.mp4`的输出mp4文件。
|
||
<p align="center">
|
||
<img src="./assets/docs/inference-animals.gif" alt="image">
|
||
</p>
|
||
|
||
#### 驱动视频自动裁剪 📢📢📢
|
||
|
||
> [!IMPORTANT]
|
||
> 使用您自己的驱动视频时,我们**推荐**: ⬇️
|
||
>
|
||
> - 将其裁剪为**1:1**的宽高比(例如,512x512或256x256像素),或通过`--flag_crop_driving_video`启用自动裁剪。
|
||
> - 专注于头部区域,类似于示例视频。
|
||
> - 最小化肩部运动。
|
||
> - 确保驱动视频的第一帧是具有**中性表情**的正面面部。
|
||
|
||
以下是通过`--flag_crop_driving_video`自动裁剪的示例:
|
||
|
||
```bash
|
||
python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d13.mp4 --flag_crop_driving_video
|
||
```
|
||
|
||
如果自动裁剪的结果不理想,您可以修改`--scale_crop_driving_video`、`--vy_ratio_crop_driving_video`选项来调整比例和偏移,或者手动进行调整。
|
||
|
||
#### 动作模板制作
|
||
|
||
您也可以使用以`.pkl`结尾的自动生成的动作模板文件来加快推理速度,并**保护隐私**,例如:
|
||
```bash
|
||
python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d5.pkl # 人像动画
|
||
python inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d5.pkl # 人像视频编辑
|
||
```
|
||
|
||
### 4. Gradio 界面 🤗
|
||
|
||
我们还提供了Gradio界面 <a href='https://github.com/gradio-app/gradio'><img src='https://img.shields.io/github/stars/gradio-app/gradio'></a>,以获得更好的体验,只需运行:
|
||
|
||
```bash
|
||
# 对于Linux和Windows用户(以及搭载Intel的macOS??)
|
||
python app.py # 人类模型模式
|
||
|
||
# 对于搭载Apple Silicon的macOS用户,不支持Intel,这可能比RTX 4090慢20倍
|
||
PYTORCH_ENABLE_MPS_FALLBACK=1 python app.py # 人类模型模式
|
||
```
|
||
|
||
我们还为动物模式提供了Gradio界面,这仅在Linux上经过NVIDIA GPU测试:
|
||
```bash
|
||
python app_animals.py # animals mode 🐱🐶
|
||
```
|
||
|
||
您可以指定`--server_port`、`--share`、`--server_name`参数以满足您的需求!
|
||
|
||
🚀我们还提供了一个加速选项`--flag_do_torch_compile`。第一次推理触发优化过程(约一分钟),使后续推理速度提高20-30%。不同CUDA版本的性能提升可能有所不同。
|
||
|
||
```bash
|
||
# 启用torch.compile以进行更快的推理
|
||
python app.py --flag_do_torch_compile
|
||
```
|
||
**注意**:此方法在Windows和macOS上不受支持。
|
||
|
||
**或者,在[HuggingFace](https://huggingface.co/spaces/KwaiVGI/LivePortrait)上轻松尝试**🤗。
|
||
|
||
### 5. 推理速度预估 🚀🚀🚀
|
||
我们还提供了一个脚本来评估每个模块的推理速度:
|
||
|
||
```bash
|
||
# 对于NVIDIA GPU
|
||
python speed.py
|
||
```
|
||
|
||
结果在[**本仓库该文件展示**](./assets/docs/speed.md).
|
||
|
||
## 社区资源 🤗
|
||
|
||
发现社区贡献的宝贵资源,以增强您的LivePortrait体验:
|
||
|
||
- [ComfyUI-LivePortraitKJ](https://github.com/kijai/ComfyUI-LivePortraitKJ) by [@kijai](https://github.com/kijai)
|
||
- [ComfyUI-AdvancedLivePortrait](https://github.com/PowerHouseMan/ComfyUI-AdvancedLivePortrait) by [@PowerHouseMan](https://github.com/PowerHouseMan).
|
||
- [comfyui-liveportrait](https://github.com/shadowcz007/comfyui-liveportrait) by [@shadowcz007](https://github.com/shadowcz007)
|
||
- [LivePortrait In ComfyUI](https://www.youtube.com/watch?v=aFcS31OWMjE) by [@Benji](https://www.youtube.com/@TheFutureThinker)
|
||
- [LivePortrait hands-on tutorial](https://www.youtube.com/watch?v=uyjSTAOY7yI) by [@AI Search](https://www.youtube.com/@theAIsearch)
|
||
- [ComfyUI tutorial](https://www.youtube.com/watch?v=8-IcDDmiUMM) by [@Sebastian Kamph](https://www.youtube.com/@sebastiankamph)
|
||
- [Replicate Playground](https://replicate.com/fofr/live-portrait) and [cog-comfyui](https://github.com/fofr/cog-comfyui) by [@fofr](https://github.com/fofr)
|
||
|
||
以及我们社区的许多其他令人惊叹的贡献!
|
||
|
||
## 致谢 💐
|
||
|
||
我们要感谢[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)和[X-Pose](https://github.com/IDEA-Research/X-Pose)仓库的的贡献者,感谢他们的开放研究和贡献。
|
||
|
||
## 引用 💖
|
||
|
||
如果您发现LivePortrait对您的研究有用,欢迎引用我们的工作,使用以下BibTeX:
|
||
|
||
```bibtex
|
||
@article{guo2024liveportrait,
|
||
title = {LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control},
|
||
author = {Guo, Jianzhu and Zhang, Dingyun and Liu, Xiaoqiang and Zhong, Zhizhou and Zhang, Yuan and Wan, Pengfei and Zhang, Di},
|
||
journal = {arXiv preprint arXiv:2407.03168},
|
||
year = {2024}
|
||
}
|
||
```
|
||
|
||
## 联系方式 📧
|
||
|
||
[**Jianzhu Guo (郭建珠)**](https://guojianzhu.com); **guojianzhu1994@gmail.com**;
|
||
|
||
## 语言
|
||
|
||
[English](./README.md) | 简体中文
|