Open Source

ESPnet (Muskits)

Overview

• This project is based on Kaldi and has been out for a while
• Recently added singing voice synthesis recipe
  • Based on Muskits framework – was merged into espnet
  • Task name is SVS: Singing Voice Synthesis
  • Various language support – Japanese,English, Korean, Chinese
• Based on ofuton_p_utagoe_db

Usage

• Colab notebook
• 2022 Tutorial at CMU

DiffSinger

DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism

Overview

• Official Torch implementation of AAAI-2022 paper
• Has two versions
  • TTS (text-to-speech) version
  • SVS (Singing Voice Synthesis) version

Usage

• TTS tutorial
• Chinese singing demo (trained on 5 hours of single female singing voice samples of Opencpop dataset)

OpenUtau

OpenUtau is an open source editing environment for UTAU community with modern user experience.

Overview

• Official Torch implementation of AAAI-2022 paper
• Has two versions
  • TTS (text-to-speech) version
  • SVS (Singing Voice Synthesis) version

Usage

• TTS tutorial