torch-tacotron

(Unofficial) PyTorch implementation of Tacotron, Wang et al., 2017.

• Tacotron: Towards End-to-End Speech Synthesis [arXiv:1703.10135]
• Predict mel-spectrogram to use other neural vocoder.

Requirements

Tested in python 3.7.9 ubuntu conda environment, requirements.txt

Usage

Download LJSpeech dataset from official:keithito

To train model, run train.py.

python train.py --data-dir /datasets/LJSpeech-1.1

Or dump the dataset to accelerate the train.

python -m utils.dump \
    --data-dir /datasets/LJSpeech-1.1 \
    --output-dir /datasets/LJSpeech-1.1/dump \
    --num-proc 8

python train.py \
    --data-dir /datasets/LJSpeech-1.1/dump \
    --from-dump

To start to train from previous checkpoint, --load-epoch is available.

python train.py \
    --data-dir /datasets/LJSpeech-1.1/dump \
    --from-dump \
    --load-epoch 20 \
    --config ./ckpt/t1.json

Checkpoint will be written on TrainConfig.ckpt, tensorboard summary on TrainConfig.log.

python train.py
tensorboard --logdir ./log

Inference and pretrained.

python inference.py \
    --config ./ckpt/t1.json \
    --ckpt ./ckpt/t1/t1_200.ckpt \
    --text "Hello, my name is revsic."

Pretrained checkpoints are relased on releases.

To use pretrained model, download files and unzip it. Followings are sample script.

from config import Config
from taco import Tacotron

with open('t1.json') as f:
    config = Config.load(json.load(f))

ckpt = torch.load('t1_200.ckpt', map_location='cpu')

tts = Tacotron(config.model)
tts.load(ckpt)

Learning Curve

Samples

Reference https://revsic.github.io/torch-tacotron.