1. fix video2video bug, recommended not to used relative_motion
2. download FasterLivePortrait-windows.zip 夸克 or FasterLivePortrait-windows.zip HuggingFace

增加JoyVASA的支持，可以用音频驱动视频或图片。非常酷！

Added support for JoyVASA, which can drive videos or images with audio. Very cool!

[中文]

1. 增加保存pickle，并支持用pickle做驱动。
2. 增加image驱动的功能，记得把relative motion取消才会有效果。
3. 增加animation_region的功能，现在可以单独控制嘴唇、眼睛等。

[English]

1. Add the ability to save pickle files and use them for driving.
2. Add the image driving feature; remember to disable relative motion for it to be effective.
3. Add the animation_region feature, allowing you to control areas like lips and eyes separately.

1. 增加animal模型的tensorrt推理。
2. 增加对mediapipe的支持，可替代insightface。
3. 用torchgeometry + cuda优化paste_back函数，现在速度提升了很多。
4. 修复Xpose的ops在一些显卡运行报错的问题等bug。

夸克网盘
需要下载最新的源代码，然后解压覆盖到windows整合包。只需要下载代码，算是一个补丁，避免又需要下载很大的windows 整合包

1. Added TensorRT inference for the animal model.
2. Added support for Mediapipe, which can replace InsightFace.
3. Optimized the paste_back function with torchgeometry and CUDA, which greatly improved the speed.
4. Fixed bugs such as errors when running Xpose's ops on some graphics cards.

Google Drive download

Please download the latest source code and then unzip it to overwrite the Windows package. You only need to download the code, which is essentially a patch to avoid having to download the large Windows package again.

1. 增加 vid2vid的支持，注意跟原版的会有点差距，我还没有集成smooth算法。
2. 删除ffmpeg的依赖。
3. 增加更多代码运行的log。

夸克网盘下载

1. Added support for vid2vid, note that there will be some differences from the original version as I have not yet integrated the smooth algorithm.
2. Removed the dependency on ffmpeg.
3. More log in running.

HF下载

Install-free, extract-and-play Windows package with TensorRT support now available.
HuggingFace:FasterLivePortrait-windows
或
夸克网盘:FasterLivePortrait-windows