CjangCjengh / vits

VITS implementation of Japanese, Chinese, Korean, Sanskrit and Thai

How to use

(Suggestion) Python == 3.7

Clone this repository

git clone https://github.com/CjangCjengh/vits.git

Choose cleaners

• Fill "text_cleaners" in config.json
• Edit text/symbols.py
• Remove unnecessary imports from text/cleaners.py

Install requirements

pip install -r requirements.txt

Create datasets

Single speaker

"n_speakers" should be 0 in config.json

path/to/XXX.wav|transcript

• Example

dataset/001.wav|こんにちは。

Mutiple speakers

Speaker id should start from 0

path/to/XXX.wav|speaker id|transcript

• Example

dataset/001.wav|0|こんにちは。

Preprocess

If you have done this, set "cleaned_text" to true in config.json

# Single speaker
python preprocess.py --text_index 1 --filelists path/to/filelist_train.txt path/to/filelist_val.txt

# Mutiple speakers
python preprocess.py --text_index 2 --filelists path/to/filelist_train.txt path/to/filelist_val.txt

Build monotonic alignment search

cd monotonic_align
python setup.py build_ext --inplace
cd ..

Train

# Single speaker
python train.py -c <config> -m <folder>

# Mutiple speakers
python train_ms.py -c <config> -m <folder>

Inference

Online

See inference.ipynb

Offline

See MoeGoe

Running in Docker

docker run -itd --gpus all --name "Container name" -e NVIDIA_DRIVER_CAPABILITIES=compute,utility -e NVIDIA_VISIBLE_DEVICES=all "Image name"