search

CSTR-Edinburgh / merlin

This is now the official location of the Merlin project.
http://www.cstr.ed.ac.uk/projects/merlin/
Apache License 2.0
1.31k stars 441 forks source link

WORLD vocoder #108

Closed malradhi closed 7 years ago

malradhi commented 7 years ago

Hi all,

How can I use WORLD vocoder alone without using any DNN training and without parameters generation? I mean just analysis and synthesis under Merlin toolkit.

Thanks in advance ....

RasmusD commented 7 years ago

Two options:

  1. For the original world - simply download it here and run it on your wavs: http://ml.cs.yamanashi.ac.jp/world/english/

  2. For the version packed with Merlin - simply run the acoustic feature extraction step on the wavs, then run the synthesis step on those extracted features.

The difference is that the original world does not natively create the kind of MGC, BAP and LF0 features extracted for Merlin DNN modelling - this is a modification performed to allow for easier modelling - so the results will be a little bit different depending on the method.

2017-03-07 11:04 GMT+00:00 Ali20177 notifications@github.com:

Hi all,

How can I use WORLD vocoder alone without using any DNN training and without parameters generation? I mean just analysis and synthesis under Merlin toolkit.

Thanks in advance ....

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/CSTR-Edinburgh/merlin/issues/108, or mute the thread https://github.com/notifications/unsubscribe-auth/AEyDp4uQvFytWdM5xiep0j6oD2jVYhNDks5rjTm8gaJpZM4MVVpv .

ronanki commented 7 years ago

You can use below scripts to do analysis-synthesis: https://github.com/CSTR-Edinburgh/merlin/tree/master/misc/scripts/vocoder/world

malradhi commented 7 years ago

Thank you very much for both of you .... It works for me.