Skip to content

Third party:Speech

Bertrand Benoit edited this page Dec 7, 2019 · 2 revisions

title: Third-party:Speech permalink: /Third-party:Speech/

TOC

Default (and recommended) mode is espeak+mbrola.

espeak mode

Since Hemera v0.1. * Speech synthesizer: eSpeak RedHat-based distributions yum install espeak Debian-based distributions

sudo apt-get install espeak

Check/update the following [configuration elements](/Appendix#Configuration_file]]

hemera.core.speech.tool.*

espeak+mbrola mode

Since Hemera v0.1. To improve quality, it is interesting to use eSpeak as Phoneme mnemonics writer and MBROLA Project as speech synthesizer.

mbrola binary

On RedHat-based distributions, mbrola is not available with package manager. To get it, follow these instructions:

  • download the binary
  • rename the wanted version to [HEMERA_TP_PATH](/Appendix#HEMERA_TP_PATH]]/speech/bin/mbrola

On Debian-based distributions, mbrola binary and data files are available with package manager: sudo apt-get install mbrola After mbrola setup, check/update the following [configuration element](/Appendix#Configuration_file]]

hemera.core.speech.mbrola.path

espeak and mbrola compatibility

Important: with espeak v1.44 and newer, you must use the new option --pho to work with mbrola; check/update the following [configuration element](/Appendix#Configuration_file]]

hemera.core.speech.tool.options

Models

You can use any model compatible with mbrola.

This is specific instructions for tested models: [English|Third-party:Speech:Models:en]], [French|Third-party:Speech:Models:fr]]

Category:HemeraBook/en Category:admin/en

You can’t perform that action at this time.