espeak, mbrola, utf8, navit : problem with french accentuated letters

[Xavier Cremaschi]

Now I am trying to plug everything into speed-dispatcher...

I added in /etc/speech-dispatcher/speechd.conf
   DefaultLanguage "fr"
   AddModule "mbrola" "sd_generic" "mbrola.conf"
   DefaultModule mbrola

and I created a mbrola.conf in /etc/speech-dispatcher/modules

It works (spd-say "bonjour" speaks!), but instead of a generic rule :
GenericExecuteSynth \
"echo \"$DATA\" \
| espeak -v mb-$VOICE $PUNCT -s $RATE -a $VOLUME \
| mbrola -v $VOLUME -e /usr/share/mbrola/$VOICE - -.au \
| $PLAY_COMMAND

I had to hardcode some variables :
GenericExecuteSynth \
"echo \"$DATA\" \
| espeak -v mb-fr1 \
| mbrola -e /usr/share/mbrola/fr1 - -.au \
| aplay
because /var/log/mbrola.log told me :
Fri Jan  2 11:34:19 2009 [765343]: child: synth command = |echo 
"bonjour" | espeak -v mb-fr1   -s 160 -a 2.00 | mbrola -v $VOLUME -e 
/usr/share/mbrola/$VOICE - -.au | $PLAY_COMMAND
|
In this line you can see that all variables have not been expanded.


Xavier.

PS : here is my mbrola.conf (with debug info enabled, cf. last line).

# Espeak mbrola output module is based on the generic plugin for Speech
# Dispatcher. It means there is no code written explicitly for
# this plugin, all the specifics are handled in this configuration
# and we call a simple command line client to perform the actual
# synthesis. Use this config file with the sd_generic output module.
#
# IMPORTANT: The audio output method relies on an audio playback
# utility (play, aplay, paplay for OSS, ALSA or Pulse)
# being installed. If this is not the case, consider installing it
# or replace the $PLAY_COMMAND string in the GenericExecuteString below
# with play, paplay or similar.
#
# GenericExecuteSynth is the shell command that should be
# executed in order to say some message. This command must
# stop saying the message on SIGKILL, otherwise it's useless.
# You can use the variables $LANGUAGE, $VOICE, $PITCH and $RATE
# which will be substituted for the appropriate value (you
# can modify this value, see other parameters).
# The command can be split into more lines, if necessary, using '\'.
#GenericExecuteSynth \
#"echo \"$DATA\" \
#| espeak -v mb-$VOICE $PUNCT -s $RATE -a $VOLUME \
#| mbrola -v $VOLUME -e /usr/share/mbrola/$VOICE - -.au \
#| $PLAY_COMMAND
GenericExecuteSynth \
"echo \"$DATA\" \
| espeak -v mb-fr1 \
| mbrola -e /usr/share/mbrola/fr1 - -.au \
| aplay

# The following three items control punctuation levels None, Some, and All.
# Each of these values will be substituted into the $PUNCT variable 
depending
# on the value passed to speech dispatcher from applications. 

# Note that if an empty string is specified, then $PUNCT will be blank 

# which is a default situation for espeak. 

 

GenericPunctNone " " 

GenericPunctSome "--punct=\"()[]{};:\"" 

GenericPunctAll "--punct" 


# GenericStripPunctChars is a list (enclosed in doublequotes) of
# all the characters that should be replaced by whitespaces in
# order not to be badly handled by the output module or misinterpreted
# by shell.
# GenericStripPunctChars  ""

# If the language you need to pass in $LANG is different
# from the standard ISO language code, you can specify
# which string to use instead. If you wish to use
# other than ISO charset for the specified language,
# you can add it's name (as accepted by iconv) as a
# third parameter in doublequotes.


GenericLanguage           "af" "af" "utf-8"
GenericLanguage           "cs" "cs" "utf-8"
GenericLanguage           "de" "de" "utf-8"
GenericLanguage           "el" "el" "utf-8"
GenericLanguage           "en" "en" "utf-8"
GenericLanguage           "es" "es" "utf-8"
GenericLanguage           "fr" "fr" "utf-8"
GenericLanguage           "hr" "hr" "utf-8"
GenericLanguage           "hu" "hu" "utf-8"
GenericLanguage           "it" "it" "utf-8"
GenericLanguage           "la" "la" "utf-8"
GenericLanguage           "nl" "nl" "utf-8"
GenericLanguage           "pl" "pl" "utf-8"
GenericLanguage           "ro" "ro" "utf-8"
GenericLanguage           "sv" "sv" "utf-8"
GenericLanguage           "sw" "sw" "utf-8"
GenericLanguage           "gr" "gr" "utf-8"
GenericLanguage           "cr" "cr" "utf-8"
GenericLanguage           "hu" "hu" "utf-8"
GenericLanguage           "id" "id" "utf-8"

# AddVoice specifies which $VOICE string should be assigned to
# each language and symbolic voice name. All the voices you want
# to use must be specified here. This list will likely not be
# up-to-date, please check eSpeak documentation and add the voices
# you want to use. Or better use the native espeak module ('espeak'
# not 'espeak-generic')

AddVoice        "en"    "MALE1"         "en1"
AddVoice        "en"    "MALE2"         "us2"
AddVoice        "en"    "MALE3"         "us3"
AddVoice        "en"    "FEMALE1"       "us1"

AddVoice        "de"    "MALE1"         "de4"
AddVoice        "de"    "MALE2"         "de6"
AddVoice        "de"    "FEMALE1"       "de5"
AddVoice        "de"    "FEMALE2"       "de7"

AddVoice        "fr"    "MALE1"       "fr1"
AddVoice        "fr"    "FEMALE1"     "fr4"

AddVoice        "it"    "MALE1"         "it3"
AddVoice        "it"    "FEMALE1"       "it4"

AddVoice        "es"    "MALE1"         "es1"
AddVoice        "es"    "MALE2"         "es2"

AddVoice        "sw"    "MALE1"         "sw1"
AddVoice        "sw"    "FEMALE1"       "sw2"

AddVoice        "gr"    "MALE1"         "gr1"
AddVoice        "gr"    "MALE2"         "gr2"

AddVoice        "cs"    "MALE1"         "cz2"
AddVoice        "pl"    "MALE1"         "pl1"
AddVoice        "ro"    "MALE1"         "ro1"
AddVoice        "af"    "MALE1"         "af1"
AddVoice        "cr"    "MALE1"         "cr1"
AddVoice        "hu"    "MALE1"         "hu1"
AddVoice        "la"    "MALE1"         "la1"
AddVoice        "id"    "MALE1"         "id1"

# These parameters set _rate_, _pitch_, and _volume_ conversion. This is
# part of the core of the definition of this generic output
# module for this concrete synthesizer, it's not intended to
# be modified by common users.
# The resulting rate (or pitch) has the form:
#       (speechd_rate * GenericRateMultiply) + GenericRateAdd
# while speechd_rate is a value between -100 (lowest) and +100 (highest)
# You have to define some meaningful conversion for each synthesizer

# Here's the mapping from SSIP (Speech Dispatcher) to ESpeak (v1.10):
#
#           SSIP Range  SSIP Default    ESpeak/MBROLA Range ESpeak Default
#           ----------- ------------    ------------         --------------
# Rate      -100 to 100     0           80 to 320       160
# Pitch     -100 to 100     0            0 to 99         50
# Volume    -100 to 100     0            0 to 2          --
#
# The SSIP defaults are actually controlled via DefaultRate, 
DefaultPitch, and
# DefaultVolume in the speechd.conf file.

GenericRateAdd          160
GenericPitchAdd         50
GenericVolumeAdd        1

# (These values are multiplied by 100, because DotConf currently
# doesn't support floats. So you can write 0.85 as 85 and so on.)

GenericRateMultiply     160
GenericPitchMultiply    50
GenericVolumeMultiply   1

# If the client program can't handle floats, you will have to
# use these two options to force integers as the parameters
# 1 means force integers, 0 means do nothing (write floats).

GenericRateForceInteger     1
GenericPitchForceInteger    1
GenericVolumeForceInteger   0

# Note that SSIP rates < -50 are spoken at -50.

# Debug turns debugging on or off
# See speechd.conf for information where debugging information is stored
Debug 1