Model Replacement 模型替换

VITS MODEL

First, follow the instructions here to convert the model into the ncnn format.

Windows

In the command line, enter:

.\convert_model.exe -c \path\to\config.json -m \path\to\model.pth -fp16

Linux

In the command line, enter:

./convert_model -c /path/to/config.json -m /path/to/model.pth -fp16

For Python Users

1. Create a virtual environment

conda create -n vits-ncnn python=3.10
conda activate vits-ncnn

1. Install dependencies

pip install -r requirements.txt

1. Run the script

python convert_model.py -c /path/to/config.json -m /path/to/model.pth -fp16

Once successful, a config.json and many .bin files will be generated. Place them all into the phone’s file manager directory:

File Management：
├─chatwaifu
     ├─ vits
         ├─ yourModel
              ├─ config.json
              ├─ xxxx.bin
              ├─ xxxx.bin
              ├─ .....

Live2D MODEL

Supports the official standard Live2D format. You need to place them in the chatwaifu/live2d/ directory within your phone’s file manager. After placing them, the directory will look like this:

File Management：
├─chatwaifu
     ├─ live2d
         ├─ yourModel
              ├─ yourModel.model3.json
              ├─ xxx.moc3
              ├─ xxx.exp3.json
              ├─ .....

The name of the Live2D model must be the same as the directory name of the VITS model, and the Live2D model3.json file must also match the directory name. If your VITS model is a multi-speaker model, you need to specify the Speaker ID in the Settings page. If not specified, the default is 0.