-
Notifications
You must be signed in to change notification settings - Fork 0
Running VTS on Linux
The following are instructions on making VTube Studio run with OpenSeeFace face tracking on Linux (specifically Arch Linux). This will require some technical knowledge about Linux and is not recommended for beginners, but if you're reading this I'm sure you're already aware of that.
Thanks to Ruyi#0110 for making this guide.
Used distro: Arch Linux. The dollar sign represents the start of the command line, don't include that.
$sudo pacman -Sy python39
Note that pacman
cannot run without sudo
. pacman
is the package manager on the used Arch install, yours might be different.
-Sy
is the argument for pacman
that tells it to update the mirrors and install the package.
python39
is the package name itself (I think on debian it might be called 3.9).
$sudo pacman -Sy python-pip python-virtualenv git
python-pip
is the thing we will use to install stuff in our virtual environment.
python-virtualenv
is the virtual environment itself.
git
will help us download OpenSeeFace
. Again, the package names might be different. Some packages you might already have installed.
$git clone https://github.com/emilianavt/OpenSeeFace
Then enter its folder:
$cd OpenSeeFace
And create the virtual environment:
$virtualenv -p python39 env
Then start the virtual environment
$source env/bin/activate
And install all the stuff the facetracker requires to run:
$pip install onnxruntime opencv-python pillow numpy==1.21.6
The reason numpy
is different is because we need an older version of numpy
. This specific version is required.
And then, this is where the magic happens!
$python facetracker.py -c 0 -W 1280 -H 720 --discard-after 0 --scan-every 0 --no-3d-adapt 1 --max-feature-updates 900
Now you should be able to open up VTube Studio, select VTubeStudioCam
and enjoy (don't worry if the resolution says 4x4, all we need is the OpenSeeFace
info from it).
Note that you need to start the virtual environment every time you want to use OpenSeeFace
, using the following three commands:
$cd OpenSeeFace
$source env/bin/activate
$python facetracker.py -c 0 -W 1280 -H 720 --discard-after 0 --scan-every 0 --no-3d-adapt 1 --max-feature-updates 900
Note that -c 0
represents your camera. It usually is 0
but if you have multiple cameras (like the camera from the Valve index) you may need to use a different number here (try 1
, 2
, ...)
如果你遇到了任何该手册中没有回答的问题,请到 VTube Studio Discord !!
- 简介&必要环境
- 已知的问题
- 哪里可以获得模型?
- Android vs. iPhone vs. 网络摄像头
- 入门指南
- 载入自定义模型
- 载入自定义背景
- VTube Studio设置
- 屏幕光源着色
- 视觉特效
- Spout2背景
- VTS模型设置
- 按键表情
- 动画
- 动画、表情、面部追踪、物理等之间的交互
- 向Mac/PC推流
- 使用OBS录制/串流
- 不通过Steam启动应用
- 以管理员身份启动
- 恢复旧的VTS版本
- NVIDIA Broadcast面部捕捉
- Google Mediapipe面部捕捉
- Tobii 眼球捕捉
- 手部捕捉
- 音频口型同步
- 使用单个网络摄像头或iPhone Android设备控制多个模型
- 连接问题&解决方案
- 截图与分享
- 挂件系统
- Live2D挂件
- Web挂件
- Web挂件插件
- 重新着色模型和物品
- 录制动画
- 多人联机
- VNet网络安全
- Twitch互动
- 添加特殊的ArtMesh功能
- 挂件场景和挂件热键
- 插件 (YouTube, Twitch等平台)
- VTube Studio模型文件
- 在模型间复制配置
- 将数据传输至VSeeFace
- Steam创意工坊
- 如何解决延迟问题
- 如何解决网络摄像头问题
- Live2D Cubism Editor 交互
- 常见问题
- 隐私政策、许可与更多条款