ICASSP 2020 ESPnet-TTS: Merlin baseline system

• Paper: https://arxiv.org/abs/1910.10909
• Demo samples: https://espnet.github.io/icassp2020-tts/

See the notebook in the repository for samples and code to reproduce.

• Direct link to the notebook: nbviewer

Note: Since audio samples are embedded in the notebook, the notebook size is quite large (~ 100 MB). It may take a while for rendering. It might be better to see the notebook locally.

Acknowledgements

Merlin is used to generate phone-level alignments. See https://github.com/r9y9/merlin/releases/tag/icassp2020-espnet-tts for code.