This is the final project of course INF573 - Image Analysis and Computer Vision. We implement a Genshin VTuber base on VTuber-MomoseHiyori by KennardWang.

原神启动！

This program is implemented base on Windows 10 x64 System. Using Python 3.8 and Unity 2019.4.1f1. The Unity Engine use port number 14514.

Make sure you have the required dependencies installed by running (Recommended to use a conda virtual environment):

pip install -r requirements.txt

• --cam: Specify the index of camera. (Default: 0)
• --debug: Whether show image and marks or not. (Default: False)
• --connect: Whether connect to unity character or not. (Default: False)
• --video: Specify the path to the video file (if not using camera). (Default: None)

Step 1 : Test camera

Run the following code, you should see your face with marks.

python main.py --debug

Press q to exit the program.

Step 2 : Connect Unity

First, run the Unity executable file Genshin-vtuber.exe.

Then run with camera

python main.py --connect

or

run with video (replace /path_of_video with your actual paths)

python main.py --connect --video /path_of_video

Press q to exit the program.

• VTuber-MomoseHiyori by KennardWang

• Head Pose Estimation by Mostafa-Nafie

• Google MediaPipe

• VTuber Live2D model by HyouKoki