• Voice Names
• Windows Installation
• Linux Installation
  • Installation of standard packages
  • Installation of MBROLA package from source
• Usage
• Adding new MBROLA voice entry to eSpeak NG
  • 1. Add MBROLA voice definition file
  • 2. Add MBROLA phoneme translation file
  • 3. Compile voice and update Makefile.am file

The MBROLA is an open source speech engine with collection of diphone voices for speech synthesis. Voices are cost-free for non-commerical purposes, but are not open source.

MBROLA does not include any text-to-phoneme translation, so this must be done by another program. eSpeak NG can be used as a front-end to MBROLA, to provide spelling-to-phoneme translation and intonation, which MBROLA then uses to generate speech sound.

To use a MBROLA voice, eSpeak NG needs information to translate from its own phonemes to the equivalent MBROLA phonemes.

The eSpeak NG voices which use MBROLA are named as:

mb-xxN

where xxN is the name of a MBROLA voice (e.g. mb-en1 for the MBROLA en1 English voice). These voice files are in eSpeak NG's folder .../espeak-ng-data/voices/mb (where ... is usually /usr/share, but could be somewhere else).

There are some additional eSpeak NG MBROLA voices, which speak English text using a MBROLA voice for a different language. These contain the name of the MBROLA voice with a suffix -en. For example, the voice mb-de4-en will speak English text with a German accent by using the MBROLA de4 voice.

The following MBROLA voices are available in eSpeak NG¹.

eSpeak NG will look for mbrola voices firstly in espeak-ng-data/mbrola and then in /usr/share/mbrola.

The installation instructions below use the MBROLA voice en1 as an example. You can use other mbrola voices for which there is an equivalent eSpeak NG voice in espeak-ng-data/voices/mb.

The SAPI5 version of eSpeak NG uses the mbrola.dll.

1. Install eSpeak NG. Include the voice mb-en1 in the list of voices during the eSpeak NG installation.

2. Install the PC/Windows version of MBROLA (MbrolaTools35.exe) from: http://www.tcts.fpms.ac.be/synthesis/mbrola/bin/pcwin/MbrolaTools35.exe.

3. Download the en1 or other voice from: https://github.com/numediart/MBROLA-voices

4. Save downloaded en1 data file into C:/Program Files/eSpeak/espeak-ng-data/mbrola.

5. Use the voice espeak-MB-EN1 from the list of SAPI5 voices.

There are standard packages prepared for MBROLA binary and voices on different Linux distributions. On Debian/Ubuntu-like Linux, you can install mbrola using apt-get package manager:

sudo apt-get install mbrola mbrola-en1

where:

• mbrola is package containing MBROLA executable,
• mbrola-en1 is mbrola data files for en1 MBROLA voice.

You can check other available voices searching with command:

apt-cache search mbrola

If some MBROLA voice is not available as Debian package, you can download data files from https://github.com/numediart/MBROLA-voices and save them in /usr/share/mbrola/xxN/xxN file, where xxN is code and number of language file.

1. Install necessary packages:

sudo apt-get install git make gcc

2. Clone MBROLA git repository:

git clone https://github.com/numediart/MBROLA.git
cd MBROLA

3. Compile MBROLA:

make

4. Install MBROLA:

sudo cp Bin/mbrola /usr/bin/mbrola

5. Download the en1 or other voice from https://github.com/numediart/MBROLA-voices

6. Copy the en1 data file to /usr/share/mbrola/en1 folder.

If you use the eSpeak NG voice such as mb-en1 then eSpeak NG will use the mbrola "en1" voice, e.g.:

espeak-ng -v mb-en1 "Hello world"

To generate mbrola phoneme data (.pho file) you can use:

espeak-ng -v mb-en1 -q --pho "Hello world"

or

espeak-ng -v mb-en1 -q --pho --phonout=out.pho "Hello world"

If you have issues with sound system integration, you can pass MBROLA's output to standard output, which then can be played by ALSA (aplay) or PulseAudio (paplay) player throug the pipe:

    espeak-ng -vmb-en1 --stdout "Hello world"|aplay

To add new MBROLA voice entry for eSpeak NG you have to:

1. add MBROLA voice definition file,
2. add MBROLA phoneme translation file,
3. update Makefile.am file and compile voice.

These steps are described in details in following sections.

eSpeak NG's voice files for MBROLA voices are in espeak-ng-data/voices/mb folder. Voice definition file is in form mb-xxN and have to contain at least this line: mbrola <voice> <translation>, e.g.

mbrola en1 en1_phtrans

Where:

• en1 is the name of the MBROLA voice.
• en1_phtrans is a translation file to convert between eSpeak NG phonemes and the equivalent MBROLA phonemes.

If language code differs between MBROLA voice and eSpeak NG language, then additional line is necessary:

    language xx

where xx is code of the language in eSpeak NG.

Binary xxN_phtrans files are kept in espeak-ng-data/mbrola_ph folder and are generated from phsource/mb/xxN text files, during MBROLA voice compilation.

Additionally MBROLA voice definition file can have other optional parameters, similar to eSpeak NG voices, which are described Voices file.

MBROLA phoneme translation files specify translations from eSpeak NG phoneme names to mbrola phoneme names.

The source phoneme translation files are in .../phsource/mbrola folder and their name is in form xxN of referenced MBROLA voice.

Note:

• eSpeak NG phonemes are referenced from voice files in phsource folder of particular language e.g. ph_english and/or general phonemes file.

• MBROLA phonemes are usualy listed in README.txt file of MBROLA voice.

Each line in the mbrola phoneme translation file contains:

<control> <espeak ph1> <espeak ph2> <percent> <mbrola ph1> [<mbrola ph2>]

• <control>
  bit 0 (+1) skip the next phoneme
  bit 1 (+2) match this and previous phoneme
  bit 2 (+4) only at the start of a word
  bit 3 (+8) don't match two phonemes across a word boundary
  bit 4 (+16) add this phoneme name as a prefix to the next phoneme name (used for de4 phoneme prefix '?')
  bit 5 (+32) only in stressed syllable
  bit 6 (+64) only at the end of a word

• <espeak ph1>
  The eSpeak NG phoneme which is to be translated to an mbrola phoneme.

• <espeak ph2>
  If this field is not NULL, then the match only occurs if this field matches the next phoneme. If control bit 1 is set, then the previous rather than the next phoneme is matched. This field may also have the following values: VWL
  matches any Vowel phoneme.

• <percent>
  If this field is zero then only one mbrola phoneme is used. If this field is non-zero, then two mbrola phonemes are used, and this value gives the percentage length of the first mbrola phoneme.

• <mbrola ph1>
  The mbrola phoneme to which the eSpeak NG phoneme is translated. This field may be NULL.

• <mbrola ph2>
  The second mbrola phoneme. This field is only used if the <percent> field is not zero.

The list is searched from start to finish, until a match is found. Therefore, a line with more specific match condition should appear before a line which matches the same eSpeak NG phoneme but with a more general condition.

Note:

You can get list (and descriptions) of defined phonemes for particular eSpeak NG language by entering command in phsource folder:

egrep "^phoneme " phonemes ph_english|cut -d$' ' -f2-|sort

where ph_english is phoneme definition for particular language

Note that ph_language file can extend or override phoneme definitions in phonemes file. Translations for all defined phonemes should be given in the mbrola phoneme translation file.

If the language's *_list or *_rules files includes rules to speak words "as English" the mbrola phoneme translation file should include rules which translate English phonemes into near equivalents, so that they can spoken by the MBROLA voice.

For latest Linux verson of mbrola you can get list of diphones for particular language by entering command:

  mbrola -d /usr/share/mbrola/en1/en1

where /usr/share/mbrola/en1/en1 is location of the voice file.

When phoneme translation source file is compiled (look at next section) espeak-ng-data/mbrola_ph/xxN_phtrans file is created.

Separate MBROLA voice can be compiled using comand:

espeak-ng --compile-mbrola=<xxN>

where xxN is MBROLA voice name.

Makefile.am is build configuration file which should be extended, to include automatic compilation of newly added MBROLA voice for eSpeak NG.

Search for mbrola: \ line in Makefile.am and add additional line for newly created MBROLA voice, e.g.:

mbrola: \
...
espeak-ng-data/mbrola_ph/xxN_phtrans \

Note that it could be that several voices share the same translation file. Then translation file is named just xx.

When Makefile.am is update, when espek-ng will be compiled, e.g. by entering command automake; make -B newly added MBROLA voice will be compiled automatically.

1: You can get list of configured entries for MBROLA synthesizer with following command in espeak-ng project folder:

cd espeak-ng-data/voices/mb/; for i in $(ls); do printf "$i "; grep name $i|awk '{print $2}'; done

2: This voice has very limited set of diphones and is usable only for reading numbers.