HMM/DNN-based Speech Synthesis System (HTS) - Download

Download †

Before downloading, you must read the license and agree to it.

DNN-HSMM †

Source code and demonstrations †

• Source code
  • github
• Data
  • English tar.gz (186 MB)
  • Japanese tar.gz (157 MB)

HTS-2.3.2 †

Demonstrations †

Attention! If you use deep neural networks with USEDNN=1 option, TensorFlow-1.4 is required in your environment.

• Speaker dependent training demo
  • English tar.bz2(178 MB)
• Speaker adaptation/adaptive training demo
  • English tar.bz2(999 MB)

HTS-2.3.1 †

Demonstrations †

Attention! If you use deep neural networks with USEDNN=1 option, TensorFlow-0.12 is required in your environment.

• Speaker dependent training demo
  • English tar.bz2(178 MB)
• Known bugs/problems/patches
  • http://hts.sp.nitech.ac.jp/hts-users/spool/2017/msg00010.html

HTS-2.3 †

Source code and demonstrations †

• README
  • HTS-2.3 txt
• Document
  • HTS-2.3 pdf
• Source code
  • HTS-2.3_for_HTK-3.4.1 tar.bz2
• Slide
  • HTS-2.3 zip (66 MB)
• Speaker dependent training demo
  • English tar.bz2 (178 MB)
  • Brazilian Portuguese tar.bz2 (108 MB)
  • Japanese tar.bz2 (126 MB)
  • Japanese song tar.bz2 (97 MB)
• Speaker adaptation/adaptive training demo
  • English tar.bz2 (999 MB)

Release archive †

You will find all the information and code available for past HTS releases in release archive page.