PyTorch implementation ofEfficiently Trainable Text-to-Speech System Based on Deep Convolutional Networks with Guided Attentionbased partially on the following projects:

• https://github.com/Kyubyong/dc_tts (audio pre processing)

• https://github.com/r9y9/deepvoice3_pytorch (data loader sampler)

Online Text-To-Speech Demo

The following notebooks are executable on https://colab.research.google.com :

• Mongolian Male Voice TTS Demo

• English Female Voice TTS Demo (LJ-Speech)

For audio samples and pretrained models, visit the above notebook links.

Training/Synthesizing English Text-To-Speech

The English TTS uses the LJ-Speech dataset.

1. Download the dataset: python dl_and_preprop_dataset.py --dataset=ljspeech

2. Train the Text2Mel model: python train-text2mel.py --dataset=ljspeech

3. Train the SSRN model: python train-ssrn.py --dataset=ljspeech

4. Synthesize sentences: python synthesize.py --dataset=ljspeech

• The WAV files are saved in the samples folder.

Training/Synthesizing Mongolian Text-To-Speech

The Mongolian text-to-speech uses 5 hours audio from the Mongolian Bible.

1. Download the dataset: python dl_and_preprop_dataset.py --dataset=mbspeech

2. Train the Text2Mel model: python train-text2mel.py --dataset=mbspeech

3. Train the SSRN model: python train-ssrn.py --dataset=mbspeech

4. Synthesize sentences: python synthesize.py --dataset=mbspeech

• The WAV files are saved in the samples folder.