Skip to content
This repository has been archived by the owner on Aug 20, 2022. It is now read-only.

Training notebook for Japanese TTS model with Tacotron 2

License

Notifications You must be signed in to change notification settings

Mildemelwe/Japanese-Tacotron-2-notebook

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Japanese Tacotron 2 notebook

Training notebook for Japanese TTS model with Tacotron 2. This repo is archived in favor of https://github.com/Mildemelwe/Non-English-Tacotron-2-Training-Notebook

Usage

This is just a basic implementation of Japanese in Tacotron 2, so if you've used Tacotron 2 before you should have no trouble. For a detailed, beginner-oriented guide, check usage.pdf.

  • Transcription should use TALQu phonetics to comply with the pretrained model in use. You may use my g2p to convert from Japanese text: https://colab.research.google.com/drive/1RQGSlx2LWl0gfokup9r-nVTeI-oe-SGs?usp=sharing
  • From my testing, around 15 minutes of audio is the minimum for proper speech. Less data may produce unwanted results such as gibberish phonemes. I did a test with 8 minutes, and many phonetic combinations could not be reproduced.

Attributions

About

Training notebook for Japanese TTS model with Tacotron 2

Topics

Resources

License

Stars

Watchers

Forks