Online Text-To-Speech Demo

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

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.

Download the dataset: python dl_and_preprop_dataset.py --dataset=ljspeech
Train the Text2Mel model: python train-text2mel.py --dataset=ljspeech
Train the SSRN model: python train-ssrn.py --dataset=ljspeech
Synthesize sentences: python synthesize.py --dataset=ljspeech
- The WAV files are saved in the samples folder.

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

Download the dataset: python dl_and_preprop_dataset.py --dataset=mbspeech
Train the Text2Mel model: python train-text2mel.py --dataset=mbspeech
Train the SSRN model: python train-ssrn.py --dataset=mbspeech
Synthesize sentences: python synthesize.py --dataset=mbspeech
- The WAV files are saved in the samples folder.