English / Japanese
Sinsy is an HMM/DNN-based singing voice synthesis system. You can generate a singing voice sample by uploading the musical score (MusicXML) to this website.
Language
Vocal
Gender param.
(-0.8 ≤ x ≤ 0.8, default: 0.55)
Vibrato intensity
(0.0 ≤ x ≤ 2.0, default: 1.0)
Pitch shift
(-24 ≤ x ≤ 24, default: 0)
Musical score (.xml)
Samples
  • Genkotsu yama no tanuki-san (in Japanese)
    • f00001j_dnn_beta5
      xml
    • f00002j_dnn_beta5
      xml
    • f01018j_dnn_beta5
      xml
    • m01083j_dnn_beta5
      xml
    • f00001j
      xml
    • f00002j
      xml
    • f00004j_beta
      xml
    • f00005j
      xml
    • m01083j
      xml
  • My grandfather's clock (in English)
    • f00002e_dnn_beta5
      xml
    • f00002e
      xml
    • m00003e_beta
      xml
  • Happy brithday to you (in Chinese Mandarin)
Options
  • Feminine/Masculine-like singing voice can be synthesized by changing the gender parameter value small/big.
  • The pitch of the synthesized singing voice can be controlled in half-tone by adjusting the pitch shift value.
  • The vibrato depth can be changed by adjusting the vibrato intensity value.
MusicXML
  • Common
    • MusicXML and compressed MusicXML are supported.
    • MusicXML shuold be UTF-8 encoding.
    • The head measure needs to start with a rest.
    • The following musical symbols are supported: tie, slur, staccato, accent, dynamics, crescendo, decrescendo, breath mark.
    • Please note that, due to server limitations, the maximum duration of any uploaded song is limited to 7 minutes for HMM vocal and 5 minutes for DNN vocal.
    • The alphabet (phonemes) input are supported. How to use is here.
  • English
    • The lyrics must be entirely written in the alphabet.
  • Japanese
    • The lyrics must be entirely written in the hiragana or katakana. Each character should be written according to its pronunciation, as shown in the example below. Please use full-width characters.
      e.g.: ``こんにちは'' → ``こんにちわ''
    • The full-width sound prolongation character ``ー'' can be used.
    • The use of the character ``っ'' is permitted.
  • Chinese (Mandarin)
    • The lyrics (pinyin and tone) must be entirely written in the ASCII alphabet and ASCII digit.
Other
Demo videos
Users videos
How to use videos
News
  • 25 Dec. 2023 [Ver. 4.4]
    Compressed MusicXML file support was added.
  • 25 Dec. 2022 [Ver. 4.3]
    Japanese vocal f00001j_dnn_beta5, f00002j_dnn_beta5, f01018j_dnn_beta5, and m01083j_dnn_beta5 and English vocal f00002e_dnn_beta5 trained by DNN-based singing voice synthesis approach were added.
  • 25 Dec. 2021 [Ver. 4.2]
    Japanese vocal f00001j_dnn_beta4, f00002j_dnn_beta4, f01018j_dnn_beta4, and m01083j_dnn_beta4 and English vocal f00002e_dnn_beta4 trained by DNN-based singing voice synthesis approach were added.
  • 10 May 2021
    Phoneme tables were modified.
  • 25 Dec. 2020 [Ver. 4.1]
    Japanese vocal f00001j_dnn_beta3 and f00002j_dnn_beta3 trained by DNN-based singing voice synthesis approach were added.
  • 11 Nov. 2018 [Ver. 4.0]
    Japanese vocal f00001j_dnn_beta2 trained by DNN-based singing voice synthesis approach was updated.
  • 25 Dec. 2017 [Ver. 3.9]
    Japanese vocal m01083j was added.
    All vocals were renamed.
  • 25 Dec. 2016 [Ver. 3.8]
    Japanese vocal f005j was added.
    Japanese vocal f001j_dnn_beta trained by DNN-based singing voice synthesis approach was added.
    Synthesized singing voice quality of f001j was increased.
  • 05 Aug. 2016
    A bug of MusicXML analyzer was fixed.
    Phoneme tables were modified.
  • 25 Dec. 2015 [Ver. 3.7]
    Chinese (Mandarin) vocal f002m was added.
    Synthesized singing voice quality was increased.
  • 20 Apr. 2015
    Reference manual was updated to input English phonemes.
  • 25 Dec. 2014 [Ver. 3.6]
    Synthesized singing voice quality was increased.
    The alphabet (phonemes) input was supported.
  • 28 Aug. 2014
    SoundCloud playlist was created.
  • 02 Apr. 2014
    Sample waveforms were added.
  • 01 Apr. 2014 [Ver. 3.5]
    Synthesized singing voice quality was increased.
  • 25 Dec. 2013 [Ver. 3.4]
    Japanese vocal f004j_beta and English vocal m003e_beta were added.
    The maximum duration of synthesized singing voice was increased.
    Synthesized singing voice quality was increased.
  • 25 Dec. 2012 [Ver. 3.3]
    English version was released.