eSpeak - Free

  • ndroid app
  • ndroid app
  • ndroid app
  • ndroid app
  • Description

  • change log

  • faq

  • Comments

  • The eSpeak Text-to-Speech program is an open source speech synthesizer that supports over 70 languages. It uses spectral formant synthesis by default which sounds robotic, but can be configured to use Klatt formant synthesis or MBROLA to give it a more natural sound. This is a mirror of the eSpeak sources with the following improvements on the master branch:

    • An autotools-based build system that supports building the phoneme data and language dictionaries from source on POSIX systems.
    • Optional support for the extended Russian and Chinese dictionaries from
    • Detection of the supported audio APIs via the configure script — no need to modify the build files.
    • The NVDA voice variants are included in the espeak-data/voices/!v directory.
    • Add support for the Oromo language (om, contributed by gude432).


    Build Dependencies

    In order to build eSpeak, you need:


    Dependency Install
    autotools sudo apt-get install make autoconf automake libtool pkg-config
    c++ compiler sudo apt-get install gcc g++
    wxWidgets sudo apt-get install libwxgtk2.8-dev
    pulseaudio sudo apt-get install libpulse-dev
    portaudio sudo apt-get install libportaudio-dev


    The espeak and espeakedit programs, along with the espeak voices, can be built via the standard autotools commands:

    $ ./
    $ ./configure --prefix=/usr
    $ make

    The configure command detects various platform differences that the espeak makefiles don't cater for (e.g. different wxWidgets version) and detect the available audio setup to use automatically.

    Audio Output Configuration

    The following configure options control which audio interfaces to use:

    Option Audio Interfaces Default
    ----with-pulseaudio PulseAudio yes
    --with-portaudio PortAudio yes
    --with-sada SADA (Solaris) no

    If pulseaudio and portaudio are both enabled and available, eSpeak will choose which one to use at runtime, trying pulseaudio first before falling back to portaudio.

    eSpeak Feature Configuration

    The following configure options control which eSpeak features are enabled:

    Option Description Default
    --with-klatt Enable Klatt formant synthesis. Yes
    --with-mbrola Enable MBROLA voice support. yes
    --with-sonic Use the sonic library to support higher WPM. yes
    --with-async Enable asynchronous commands. Yes

    Extended Dictionary Configuration

    The following configure options control which of the extended dictionary files to build:

    Option Extended Dictionary Default
    --with-extdict-ru Russian no
    --with-extdict-zh Mandarin Chinese no
    --with-extdict-zhy Cantonese no

    If the option has the value no, the extended dictionary is not built; instead, the version from upstream espeak is build. If the option has the value yes, the extended dictionary is built; that is, the extra data from is used.


    Before installing, you can test the built espeak using the following command:

    	$ ESPEAK_DATA_PATH=`pwd` LD_LIBRARY_PATH=src:${LD_LIBRARY_PATH} src/espeak ...

    The ESPEAK_DATA_PATH variable needs to be set to use the espeak data from the source tree. Otherwise, espeak will look in $(HOME) or /usr/share/espeak-data. The LD_LIBRARY_PATH is set as espeak uses the shared library. This ensures that espeak uses the built shared library in the src directory and not the one on the system (which could be an older version).


    You can install eSpeak by running the following command:

    	$ sudo make LIBDIR=/usr/lib/x86_64-linux-gnu install

    The LIBDIR path may be different to the one on your system (the above is for 64-bit Debian/Ubuntu releases that use the multi-arch package structure -- that is, Debian Wheezy or later).

    You can find out where espeak is installed to on your system if you already have an espeak install by running:

    	$ find /usr/lib | grep libespeak

    Building Voices

    If you are modifying a language's phoneme, voice or dictionary files, you can just build that voice by running:

    	$ make 

    For example, if you add entries in the dictsource/en_extra file, you can run $ make en to build an English voice file with those changes in without rebuilding all the voices. This will make it easier to spot errors. :

    Adding New Voices

    Once you have added the necessary files to eSpeak to support the new voice, you can then run:

    	$ ./mkdictlist

    This will update the build system so that make will build the new voice in addition to building everything else, and add a target for building just that voice.

    Branching Structure and Changes

    The following sections describe the different branches in this repository, what they are used for and how they relate to the upstream version of eSpeak (including details of any changes made to eSpeak).
    Changes are merged in the following order:

    The upstream branch contains the unmodified eSpeak sources imported from the espeak subversion repository using the git svn command.

    In order to synchronize with the SVN repository, you need to initialize git-svn:

    	$ git svn init

    Next, you need to make git-svn look for commits on the upstream branch:

    	$ git update-ref refs/remotes/git-svn refs/remotes/origin/upstream

    The eSpeak SVN repository moved, so you need to inform git-svn to rewrite the URLs to the old location:

    	$ git config svn-remote.svn.rewriteRoot

    Now you can update from SVN by running:

    	$ git checkout upstream && git svn rebase

    The development branch replaces the code from subversion with the code from It combines the eSpeak and espeakedit zip files containing the Linux binaries with source code to match the subversion code. The motivation for using these development releases is that the source code from the subversion repositories is often incomplete -- either missing several files, or not including certain changes. This is also why the master branch pulls changes from development, not upstream.

    The following changes are performed when merging the two zip files:

    The following files are removed (excluded via .gitignore) from the sources:

    The praat-mod directory in the development zip files provides modifications of the praat program to support generating spectral data for use with espeakedit. This is not available in the subversion releases.

    Instead of adding the praat-mod folder to the sources here, to better track the changes to the praat program the modifications are included in the espeak branch of the praat mirror. This mirror currently includes support upto version 5.3.23 of praat.

    The master branch contains changes to the build system, espeakedit and bug fixes detailed in the sections below. In addition, it:

    The build system for eSpeak has been changed on the master branch to use autotools and to support building the phoneme data and language dictionaries on POSIX systems.

    The following changes have been made to espeakedit to make it usable from the command line, especially for use in the build process:

    The following bugs have been fixed on the master branch:

    The bug fixes are sent upstream for inclusion in the upstream version of eSpeak.

    The android branch contains the sources for the eSpeak for Android program, based on the eyes-free port of eSpeak to Android.

    • a functional autotools system (make, autoconf, automake, libtool and pkg-config);
    • a functional c++ compiler;
    • wxWidgets development libraries (needed to build and run espeakedit to compile the phoneme data).

    • Optionally, you need:

      • the pulseaudio development library to enable pulseaudio output;
      • the portaudio development library to enable portaudio output.
    • NOTE:  upstream => development => master => android upstream Setting Up git-svn development
      • the src/Makefile in the espeakedit code has been renamed to src/Makefile.espeakedit to match the name in subversion;
      • the espeakedit files have been merged into the eSpeak files as the subversion repository does not split out the two;
      • the platforms/riscos/Makefile,fe1 file is moved to platforms/riscos/Makefile;
      • the phsource/vnasal/.directory file is removed — this is used by the Dolphin file browser and should not be included in the sources;
      • the docs folder contains an older version of the documentation, so these changes are reverted; ;
      • the espeak-data/voices/asia/ko file is missing from the development sources, but none of the other related Korean voice files have been removed so the file is restored to the upstream version; ;
      • the executable bit is removed from the source files — this is because there are no files that need to be executable in the source tree. ;
      • the Linux binaries (including the voice/phoneme data) — these are not part of the source code; ;
      • generated files such as dictsource/dict_phonemes — these are not part of the source code; ;
      • praat-mod — this is not included as part of the subversion code (see Praat Modifications for where to get these changes); ;
      • Makefile — this is a copy of src/Makefile; ;
      • ReadMe — this is not provided in the subversion codebase, instead a more informative version is located in this file; ;
      • ChangeLog.txt — this is not provided in the subversion codebase, is incomplete and there is a more detailed change log available in the subversion and git repositories; ;
      • License.txt — this is a copy of the GPLv3 license; it is not provided in the subversion repository; for this git mirror, the GPLv3 license is located in the COPYING file as per GNU guidelines. ;
    • Praat Modifications master
      • includes the NVDA voice variants in the espeak-data/voices/!v directory;
      • supports the Oromo language (om, contributed by gude432).
      • Optional support for the extended Russian (ru), Mandarin Chinese (zh) and Cantonese Chinese (zhy) dictionaries from
      • Use /i/ instead of /ɪ/ for en transcription of the /i/ phoneme; en-n still uses the accent-specific /ɪ/ transcription.
    • Build System
      • Use autotools (autoconf, automake and libtool) to build espeak.
      • Use the portaudio header on the build machine, not the one in espeak. This avoids the need to link to the correct portaudio header file.
      • Detect wxWidgets using wx-config.
      • Add configure options and detection for the pulse, portaudio, runtime and sada audio API modes.
      • Allow CXXFLAGS to be overwritten from the environment. This is based on the patch for Debian bug #707925 by Jason White.
      • Klatt, MBROLA, libsonic and asynchronous command support was made optional. This covers:
        • adding an option to the configure script;
        • only building the source files that are enabled;
        • removing the guards around the relevant source files;
        • removing the relevant conditional compilation defines from speech.h;
        • fixing the sources if needed to build without support for the feature.
      • espeakedit
        • Build the intonation data as well as the phoneme data when the --compile option is passed.Don't compile the dictionaries when using the --compile command.
        • Only call DrawEnvelopes if MAKE_ENVELOPES is defined. This prevents the phsource/envelopes.png file being updated every build.
        • Don't load the phoneme data before compiling it, as the phoneme data may not exist.
        • Use the path set by ESPEAK_DATA_PATH if that environment variable exists, otherwise use the HOME environment variable.
        • Don't use the paths from the configuration file when running the --compile command.
      • Bug Fixes
        • all bug fixes have been applied upstream.

    There is no Features content
  • There is no Related Products
  • There is no Change Log content
  • There is no FAQ