Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #641 Questions about espeak-ng
By valdisvi:

  1. espeak-ng works only as a command line application. Adding GUI to that will just hide things and add clutter.
  2. Default voice/language for espeak-ng is en which is short-hand of en-gb. You have to pass value for -v parameter e.g. -v en-us to use different voice/language. espeak-ng --voices command shows only natively synthesized voices. You can get list of MBROLA voices here.
  3. Most of MBROLA voices are created with 16kHz sampling frequency. So, theoretical limit is 8kHz, but practically it is even lower than that (~4hHz). If there are several versions of MBROLA voices available for language, one with bigger number usually has better quality. You can only adjust phoneme translation with parameters.
  4. You can create your own MBROLA voice by recording samples, cutting diphones from them and converting samples with MBROLATOR into MBROLA database. It requires at least several months for single person to get something usable and probably more than year to get better sound than for existing voices.


[espeak-ng:master] New Comment on Issue #641 Questions about espeak-ng
By valdisvi:

  1. espeak-ng works only as a command line application. Adding GUI to that will just hide things and add clutter.
  2. Default voice/language for espeak-ng is en which is short-hand of en-gb. You have to pass value for -v parameter e.g. -v en-us to use different voice/language. espeak-ng --voices command shows only natively synthesized voices. You can get list of MBROLA voices here.
  3. Most of MBROLA voices are created with 16kHz sampling frequency. So, theoretical limit is 8kHz, but practically it is even lower than that (~4kHz). If there are several versions of MBROLA voices available for language, one with bigger number usually has better quality. You can only adjust phoneme translation with parameters.
  4. You can create your own MBROLA voice by recording samples, cutting diphones from them and converting samples with MBROLATOR into MBROLA database. It requires at least several months for single person to get something usable and probably more than year to get better sound than for existing voices.


[espeak-ng:master] New Comment on Issue #646 Suggestions: Wanted Learning ToMake Your own voice Tutorial
By valdisvi:

Short guide to create MBROLA voice is described in this project: https://github.com/numediart/MBROLATOR.

Join espeak-ng@groups.io to automatically receive all group messages.