vits-finetuning/README.md

VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech

Jaehyeon Kim, Jungil Kong, and Juhee Son

In our recent paper, we propose VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech.

Several recent end-to-end text-to-speech (TTS) models enabling single-stage training and parallel sampling have been proposed, but their sample quality does not match that of two-stage TTS systems. In this work, we present a parallel end-to-end TTS method that generates more natural sounding audio than current two-stage models. Our method adopts variational inference augmented with normalizing flows and an adversarial training process, which improves the expressive power of generative modeling. We also propose a stochastic duration predictor to synthesize speech with diverse rhythms from input text. With the uncertainty modeling over latent variables and the stochastic duration predictor, our method expresses the natural one-to-many relationship in which a text input can be spoken in multiple ways with different pitches and rhythms. A subjective human evaluation (mean opinion score, or MOS) on the LJ Speech, a single speaker dataset, shows that our method outperforms the best publicly available TTS systems and achieves a MOS comparable to ground truth.

Visit our demo for audio samples.

We also provide the pretrained models.

** Update note: Thanks to Rishikesh (ऋषिकेश), our interactive TTS demo is now available on Colab Notebook.

VITS at training	VITS at inference

Pre-requisites

0. Python >= 3.6
1. Clone this repository
2. Install python requirements. Please refer requirements.txt
   1. You may need to install espeak first: apt-get install espeak
3. Download datasets
   1. Download and extract the LJ Speech dataset, then rename or create a link to the dataset folder: ln -s /path/to/LJSpeech-1.1/wavs DUMMY1
   2. For mult-speaker setting, download and extract the VCTK dataset, and downsample wav files to 22050 Hz. Then rename or create a link to the dataset folder: ln -s /path/to/VCTK-Corpus/downsampled_wavs DUMMY2
4. Build Monotonic Alignment Search and run preprocessing if you use your own datasets.

# Cython-version Monotonoic Alignment Search
cd monotonic_align
python setup.py build_ext --inplace

# Preprocessing (g2p) for your own datasets. Preprocessed phonemes for LJ Speech and VCTK have been already provided.
# python preprocess.py --text_index 1 --filelists filelists/ljs_audio_text_train_filelist.txt filelists/ljs_audio_text_val_filelist.txt filelists/ljs_audio_text_test_filelist.txt 
# python preprocess.py --text_index 2 --filelists filelists/vctk_audio_sid_text_train_filelist.txt filelists/vctk_audio_sid_text_val_filelist.txt filelists/vctk_audio_sid_text_test_filelist.txt

Training Exmaple

# LJ Speech
python train.py -c configs/ljs_base.json -m ljs_base

# VCTK
python train_ms.py -c configs/vctk_base.json -m vctk_base

Exporting to TorchScript

Device can be either cpu or cuda. They must match at inference time, cpu-exported model on CPU and CUDA-exported model on GPU.

python export_ts.py --checkpoint vits.pth --device cpu --config ./configs/ljs_li44_tmbert_nmp_s1.json --out-path my_tsvits.pt --test-string "I like pizza"

Inference with TorchScript

Device can be either cpu or cuda. Again, this must match the device you exported to TS with, cpu-exported model on CPU and CUDA-exported model on GPU.

python infer_ts.py --out-path res --device cpu --config ./configs/ljs_li44_tmbert_nmp_s1.json --checkpoint my_tsvits.pt --text "I like hamburgers"

Check tsvitsfe.py for the simple inference frontend class. The script above just calls it and saves to a wave file

Inference Example

See inference.ipynb