/vits-cantonese

Cantonese Text to Speech with VITS implementation

Primary LanguagePythonMIT LicenseMIT

Demo outputs:

  1. 我愛你
audio.mp4

Instructions

  1. Create a virtual environment python -m venv ./venv

  2. open demo.ipynb and install the dependencies

  3. move files to opencc lib after installing opencc

    1. mv ./opencc/jyutjyu.json ./venv/lib/site-packages/opencc/clib/share/opencc/jyutjyu.json
    2. mv ./opencc/jyutjyu.ocd2 ./venv/lib/site-packages/opencc/clib/share/opencc/jyutjyu.ocd2

  4. build cpython codes cd vits/monotonic_align python setup.py build_ext --inplace

  5. download models

    1. https://huggingface.co/xiaomaiiwn/vits-cantonese/blob/main/model/G.pth

Acknowledgement

I do not own the models and you should follow the license of the models.