class SynthDriver(synthDriverHandler.SynthDriver):
supportedSettings=(SynthDriver.VoiceSetting(), SynthDriver.VariantSetting(),
SynthDriver.RateSetting(), BooleanSynthSetting("rateBoost", _("Rate boos&t")),
SynthDriver.PitchSetting(), SynthDriver.InflectionSetting(), SynthDriver.VolumeSetting(), NumericSynthSetting("hsz", _("Head Size"), False), NumericSynthSetting("rgh", _("Roughness"), False), NumericSynthSetting("bth", _("Breathiness"), False), BooleanSynthSetting("backquoteVoiceTags", _("Enable backquote voice &tags"), False))
description='IBMTTS'
name='ibmeci'
speakingLanguage=""
@classmethod
def check(cls):
return _ibmeci.eciCheck()
def __init__(self):
_ibmeci.initialize()
# This information doesn't really need to be displayed, and makes IBMTTS unusable if the addon is not in the same drive as NVDA executable.
# But display it only on debug mode in case of it can be useful
log.debug("Using IBMTTS version %s" % _ibmeci.eciVersion())
lang = languageHandler.getLanguage()
self.rate=50
self.speakingLanguage=lang
self.variant="1"
def speak(self,speechSequence):
last = None
defaultLanguage=self.language
outlist = []
for item in speechSequence:
if isinstance(item, string_types):
s = self.processText(unicode(item))
outlist.append((_ibmeci.speak, (s,)))
last = s
elif isinstance(item,speech.IndexCommand):
outlist.append((_ibmeci.index, (item.index,)))
elif isinstance(item,speech.LangChangeCommand):
l=None
if item.lang in langsAnnotations: l = langsAnnotations[item.lang]
elif item.lang and item.lang[0:2] in langsAnnotations: l = langsAnnotations[item.lang[0:2]]
if l:
if item.lang != self.speakingLanguage and item.lang != self.speakingLanguage[0:2]:
outlist.append((_ibmeci.speak, (l,)))
self.speakingLanguage=item.lang
else:
outlist.append((_ibmeci.speak, (langsAnnotations[defaultLanguage],)))
self.speakingLanguage = defaultLanguage
elif isinstance(item,speech.CharacterModeCommand):
outlist.append((_ibmeci.speak, (b"`ts1" if item.state else "b`ts0",)))
elif isinstance(item,speech.SpeechCommand):
log.debugWarning("Unsupported speech command: %s"%item)
else:
log.error("Unknown speech: %s"%item)
if last is not None and not last[-1] in punctuation: outlist.append((_ibmeci.speak, (b'`p1',)))
outlist.append((_ibmeci.setEndStringMark, ()))
outlist.append((_ibmeci.speak, (b"`ts0",)))
outlist.append((_ibmeci.synth, ()))
_ibmeci.synthQueue.put(outlist)
_ibmeci.process()
def processText(self,text):
text = text.rstrip()
if _ibmeci.params[9] in (65536, 65537): text = resub(english_fixes, text)
if _ibmeci.params[9] in (131072, 131073): text = resub(spanish_fixes, text)
if _ibmeci.params[9] in (196609, 196608):
text = resub(french_fixes, text)
text = text.replace('quil', 'qil') #Sometimes this string make everything buggy with IBMTTS in French
#if not self._backquoteVoiceTags: text = text.replace(u'‵', ' ')
if self._backquoteVoiceTags:
text = "`pp0 `vv%d %s" % (self.getVParam(_ibmeci.vlm), text.replace('`', ' ')) #no embedded commands
text = resub(anticrash_res, text)
#this converts to ansi for anticrash. If this breaks with foreign langs, we can remove it.
text = text.encode('mbcs', 'replace')
else:
#this converts to ansi for anticrash. If this breaks with foreign langs, we can remove it.
text = text.encode('mbcs', 'replace')
text = resub(anticrash_res, text)
text = b"`pp0 `vv%d %s" % (self.getVParam(_ibmeci.vlm), text.replace(b'`', b' ')) #no embedded commands
text = pause_re.sub(br'\1 `p1\2\3', text)
text = time_re.sub(br'\1:\2 \3', text)
# temporal fix: replace , with `" -" because IBMTTS seems ignore commas at the end.
# if you know a better solution please let me know to update it.
if text[-1] == b",": text = text[0:-1]+b" -"
return text
def pause(self,switch):
_ibmeci.pause(switch)
def terminate(self):
_ibmeci.terminate()
_backquoteVoiceTags=False
def _get_backquoteVoiceTags(self):
return self._backquoteVoiceTags
def _set_backquoteVoiceTags(self, enable):
if enable == self._backquoteVoiceTags:
return
self._backquoteVoiceTags = enable
_rateBoost = False
RATE_BOOST_MULTIPLIER = 1.6
def _get_rateBoost(self):
return self._rateBoost
def _set_rateBoost(self, enable):
if enable != self._rateBoost:
rate = self.rate
self._rateBoost = enable
self.rate = rate
def _get_rate(self):
val = self.getVParam(_ibmeci.rate)
if self._rateBoost: val=int(round(val/self.RATE_BOOST_MULTIPLIER))
return self._paramToPercent(val, minRate, maxRate)
def _set_rate(self,vl):
val = self._percentToParam(vl, minRate, maxRate)
if self._rateBoost: val = int(round(val *self.RATE_BOOST_MULTIPLIER))
self._rate = val
self.setVParam(_ibmeci.rate, val)
def _get_pitch(self):
return self.getVParam(_ibmeci.pitch)
def _set_pitch(self,vl):
self.setVParam(_ibmeci.pitch,vl)
def _get_volume(self):
return self.getVParam(_ibmeci.vlm)
def _set_volume(self,vl):
self.setVParam(_ibmeci.vlm,int(vl))
def _set_inflection(self,vl):
vl = int(vl)
self.setVParam(_ibmeci.fluctuation,vl)
def _get_inflection(self):
return self.getVParam(_ibmeci.fluctuation)
def _set_hsz(self,vl):
vl = int(vl)
self.setVParam(_ibmeci.hsz,vl)
def _get_hsz(self):
return self.getVParam(_ibmeci.hsz)
def _set_rgh(self,vl):
vl = int(vl)
self.setVParam(_ibmeci.rgh,vl)
def _get_rgh(self):
return self.getVParam(_ibmeci.rgh)
def _set_bth(self,vl):
vl = int(vl)
self.setVParam(_ibmeci.bth,vl)
def _get_bth(self):
return self.getVParam(_ibmeci.bth)
def _getAvailableVoices(self):
o = OrderedDict()
for name in os.listdir(_ibmeci.ttsPath):
if name.lower().endswith('.syn'):
info = _ibmeci.langs[name.lower()[:3]]
o[str(info[0])] = VoiceInfo(str(info[0]), info[1], info[2])
return o
def _get_voice(self):
return str(_ibmeci.params[9])
def _set_voice(self,vl):
_ibmeci.set_voice(vl)
def getVParam(self,pr):
return _ibmeci.getVParam(pr)
def setVParam(self, pr,vl):
_ibmeci.setVParam(pr, vl)
def _get_lastIndex(self):
#fix?
return _ibmeci.lastindex
def cancel(self):
_ibmeci.stop()
def _getAvailableVariants(self):
global variants
return OrderedDict((str(id), synthDriverHandler.VoiceInfo(str(id), name)) for id, name in variants.items())
def _set_variant(self, v):
global variants
self._variant = v if int(v) in variants else "1"
_ibmeci.setVariant(int(v))
self.setVParam(_ibmeci.rate, self._rate)
# if 'ibmtts' in config.conf['speech']:
# config.conf['speech']['ibmtts']['pitch'] = self.pitch
def _get_variant(self): return self._variant
class SynthDriver(SynthDriver):
name = "espeak"
description = "eSpeak NG"
supportedSettings=(
SynthDriver.VoiceSetting(),
SynthDriver.VariantSetting(),
SynthDriver.RateSetting(),
# Translators: This is the name of the rate boost voice toggle
# which further increases the speaking rate when enabled.
BooleanSynthSetting("rateBoost",_("Rate boos&t")),
SynthDriver.PitchSetting(),
SynthDriver.InflectionSetting(),
SynthDriver.VolumeSetting(),
)
@classmethod
def check(cls):
return True
def __init__(self):
_espeak.initialize()
log.info("Using eSpeak version %s" % _espeak.info())
lang=languageHandler.getLanguage()
_espeak.setVoiceByLanguage(lang)
self._language=lang
self._variantDict=_espeak.getVariantDict()
self.variant="max"
self.rate=30
self.pitch=40
self.inflection=75
def _get_language(self):
return self._language
PROSODY_ATTRS = {
speech.PitchCommand: "pitch",
speech.VolumeCommand: "volume",
speech.RateCommand: "rate",
}
IPA_TO_ESPEAK = {
u"θ": u"T",
u"s": u"s",
u"ˈ": u"'",
}
def _processText(self, text):
text = unicode(text)
# We need to make several replacements.
return text.translate({
0x1: None, # used for embedded commands
0x3C: u"<", # <: because of XML
0x3E: u">", # >: because of XML
0x5B: u" [", # [: [[ indicates phonemes
})
def speak(self,speechSequence):
defaultLanguage=self._language
textList=[]
langChanged=False
prosody={}
# We output malformed XML, as we might close an outer tag after opening an inner one; e.g.
# <voice><prosody></voice></prosody>.
# However, eSpeak doesn't seem to mind.
for item in speechSequence:
if isinstance(item,basestring):
textList.append(self._processText(item))
elif isinstance(item,speech.IndexCommand):
textList.append("<mark name=\"%d\" />"%item.index)
elif isinstance(item,speech.CharacterModeCommand):
textList.append("<say-as interpret-as=\"characters\">" if item.state else "</say-as>")
elif isinstance(item,speech.LangChangeCommand):
if langChanged:
textList.append("</voice>")
textList.append("<voice xml:lang=\"%s\">"%(item.lang if item.lang else defaultLanguage).replace('_','-'))
langChanged=True
elif isinstance(item,speech.BreakCommand):
textList.append('<break time="%dms" />' % item.time)
elif type(item) in self.PROSODY_ATTRS:
if prosody:
# Close previous prosody tag.
textList.append("</prosody>")
attr=self.PROSODY_ATTRS[type(item)]
if item.multiplier==1:
# Returning to normal.
try:
del prosody[attr]
except KeyError:
pass
else:
prosody[attr]=int(item.multiplier* 100)
if not prosody:
continue
textList.append("<prosody")
for attr,val in prosody.iteritems():
textList.append(' %s="%d%%"'%(attr,val))
textList.append(">")
elif isinstance(item,speech.PhonemeCommand):
# We can't use unicode.translate because we want to reject unknown characters.
try:
phonemes="".join([self.IPA_TO_ESPEAK[char] for char in item.ipa])
# There needs to be a space after the phoneme command.
# Otherwise, eSpeak will announce a subsequent SSML tag instead of processing it.
textList.append(u"[[%s]] "%phonemes)
except KeyError:
log.debugWarning("Unknown character in IPA string: %s"%item.ipa)
if item.text:
textList.append(self._processText(item.text))
elif isinstance(item,speech.SpeechCommand):
log.debugWarning("Unsupported speech command: %s"%item)
else:
log.error("Unknown speech: %s"%item)
# Close any open tags.
if langChanged:
textList.append("</voice>")
if prosody:
textList.append("</prosody>")
text=u"".join(textList)
_espeak.speak(text)
def cancel(self):
_espeak.stop()
def pause(self,switch):
_espeak.pause(switch)
_rateBoost = False
RATE_BOOST_MULTIPLIER = 3
def _get_rateBoost(self):
return self._rateBoost
def _set_rateBoost(self, enable):
if enable == self._rateBoost:
return
rate = self.rate
self._rateBoost = enable
self.rate = rate
def _get_rate(self):
val=_espeak.getParameter(_espeak.espeakRATE,1)
if self._rateBoost:
val=int(val/self.RATE_BOOST_MULTIPLIER)
return self._paramToPercent(val,_espeak.minRate,_espeak.maxRate)
def _set_rate(self,rate):
val=self._percentToParam(rate, _espeak.minRate, _espeak.maxRate)
if self._rateBoost:
val=int(val*self.RATE_BOOST_MULTIPLIER)
_espeak.setParameter(_espeak.espeakRATE,val,0)
def _get_pitch(self):
val=_espeak.getParameter(_espeak.espeakPITCH,1)
return self._paramToPercent(val,_espeak.minPitch,_espeak.maxPitch)
def _set_pitch(self,pitch):
val=self._percentToParam(pitch, _espeak.minPitch, _espeak.maxPitch)
_espeak.setParameter(_espeak.espeakPITCH,val,0)
def _get_inflection(self):
val=_espeak.getParameter(_espeak.espeakRANGE,1)
return self._paramToPercent(val,_espeak.minPitch,_espeak.maxPitch)
def _set_inflection(self,val):
val=self._percentToParam(val, _espeak.minPitch, _espeak.maxPitch)
_espeak.setParameter(_espeak.espeakRANGE,val,0)
def _get_volume(self):
return _espeak.getParameter(_espeak.espeakVOLUME,1)
def _set_volume(self,volume):
_espeak.setParameter(_espeak.espeakVOLUME,volume,0)
def _getAvailableVoices(self):
voices=OrderedDict()
for v in _espeak.getVoiceList():
l=v.languages[1:]
# #5783: For backwards compatibility, voice identifies should always be lowercase
identifier=os.path.basename(v.identifier).lower()
voices[identifier]=VoiceInfo(identifier,v.name,l)
return voices
def _get_voice(self):
curVoice=getattr(self,'_voice',None)
if curVoice: return curVoice
curVoice = _espeak.getCurrentVoice()
if not curVoice:
return ""
# #5783: For backwards compatibility, voice identifies should always be lowercase
return curVoice.identifier.split('+')[0].lower()
def _set_voice(self, identifier):
if not identifier:
return
# #5783: For backwards compatibility, voice identifies should always be lowercase
identifier=identifier.lower()
if "\\" in identifier:
identifier=os.path.basename(identifier)
self._voice=identifier
try:
_espeak.setVoiceAndVariant(voice=identifier,variant=self._variant)
except:
self._voice=None
raise
self._language=super(SynthDriver,self).language
def _get_lastIndex(self):
return _espeak.lastIndex
def terminate(self):
_espeak.terminate()
def _get_variant(self):
return self._variant
def _set_variant(self,val):
self._variant = val if val in self._variantDict else "max"
_espeak.setVoiceAndVariant(variant=self._variant)
def _getAvailableVariants(self):
return OrderedDict((ID,VoiceInfo(ID, name)) for ID, name in self._variantDict.iteritems())
class SynthDriver(SynthDriver):
name = "espeak"
description = "eSpeak"
supportedSettings = (
SynthDriver.VoiceSetting(),
SynthDriver.VariantSetting(),
SynthDriver.RateSetting(),
# Translators: This is the name of the rate boost voice toggle
# which further increases the speaking rate when enabled.
BooleanSynthSetting("rateBoost", _("Rate boos&t")),
SynthDriver.PitchSetting(),
SynthDriver.InflectionSetting(),
SynthDriver.VolumeSetting(),
)
@classmethod
def check(cls):
return True
def __init__(self):
_espeak.initialize()
log.info("Using eSpeak version %s" % _espeak.info())
lang = languageHandler.getLanguage()
_espeak.setVoiceByLanguage(lang)
self._language = lang
self._variantDict = _espeak.getVariantDict()
self.variant = "max"
self.rate = 30
self.pitch = 40
self.inflection = 75
def _get_language(self):
return self._language
def speak(self, speechSequence):
defaultLanguage = self._language
textList = []
langChanged = False
for item in speechSequence:
if isinstance(item, basestring):
s = unicode(item)
# Replace \01, as this is used for embedded commands.
#Also replace < and > as espeak handles xml
s.translate({
ord(u'\01'): None,
ord(u'<'): u'<',
ord(u'>'): u'>'
})
textList.append(s)
elif isinstance(item, speech.IndexCommand):
textList.append("<mark name=\"%d\" />" % item.index)
elif isinstance(item, speech.CharacterModeCommand):
textList.append("<say-as interpret-as=\"characters\">" if item.
state else "</say-as>")
elif isinstance(item, speech.LangChangeCommand):
if langChanged:
textList.append("</voice>")
textList.append(
"<voice xml:lang=\"%s\">" %
(item.lang if item.lang else defaultLanguage).replace(
'_', '-'))
langChanged = True
elif isinstance(item, speech.SpeechCommand):
log.debugWarning("Unsupported speech command: %s" % item)
else:
log.error("Unknown speech: %s" % item)
if langChanged:
textList.append("</voice>")
text = u"".join(textList)
_espeak.speak(text)
def cancel(self):
_espeak.stop()
def pause(self, switch):
_espeak.pause(switch)
_rateBoost = False
RATE_BOOST_MULTIPLIER = 3
def _get_rateBoost(self):
return self._rateBoost
def _set_rateBoost(self, enable):
if enable == self._rateBoost:
return
rate = self.rate
self._rateBoost = enable
self.rate = rate
def _get_rate(self):
val = _espeak.getParameter(_espeak.espeakRATE, 1)
if self._rateBoost:
val = int(val / self.RATE_BOOST_MULTIPLIER)
return self._paramToPercent(val, _espeak.minRate, _espeak.maxRate)
def _set_rate(self, rate):
val = self._percentToParam(rate, _espeak.minRate, _espeak.maxRate)
if self._rateBoost:
val = int(val * self.RATE_BOOST_MULTIPLIER)
_espeak.setParameter(_espeak.espeakRATE, val, 0)
def _get_pitch(self):
val = _espeak.getParameter(_espeak.espeakPITCH, 1)
return self._paramToPercent(val, _espeak.minPitch, _espeak.maxPitch)
def _set_pitch(self, pitch):
val = self._percentToParam(pitch, _espeak.minPitch, _espeak.maxPitch)
_espeak.setParameter(_espeak.espeakPITCH, val, 0)
def _get_inflection(self):
val = _espeak.getParameter(_espeak.espeakRANGE, 1)
return self._paramToPercent(val, _espeak.minPitch, _espeak.maxPitch)
def _set_inflection(self, val):
val = self._percentToParam(val, _espeak.minPitch, _espeak.maxPitch)
_espeak.setParameter(_espeak.espeakRANGE, val, 0)
def _get_volume(self):
return _espeak.getParameter(_espeak.espeakVOLUME, 1)
def _set_volume(self, volume):
_espeak.setParameter(_espeak.espeakVOLUME, volume, 0)
def _getAvailableVoices(self):
voices = OrderedDict()
for v in _espeak.getVoiceList():
l = v.languages[1:]
identifier = os.path.basename(v.identifier)
voices[identifier] = VoiceInfo(identifier, v.name, l)
return voices
def _get_voice(self):
curVoice = getattr(self, '_voice', None)
if curVoice: return curVoice
curVoice = _espeak.getCurrentVoice()
if not curVoice:
return ""
return curVoice.identifier.split('+')[0]
def _set_voice(self, identifier):
if not identifier:
return
if "\\" in identifier:
identifier = os.path.basename(identifier)
self._voice = identifier
try:
_espeak.setVoiceAndVariant(voice=identifier, variant=self._variant)
except:
self._voice = None
raise
self._language = super(SynthDriver, self).language
def _get_lastIndex(self):
return _espeak.lastIndex
def terminate(self):
_espeak.terminate()
def _get_variant(self):
return self._variant
def _set_variant(self, val):
self._variant = val if val in self._variantDict else "max"
_espeak.setVoiceAndVariant(variant=self._variant)
def _getAvailableVariants(self):
return OrderedDict((ID, VoiceInfo(ID, name))
for ID, name in self._variantDict.iteritems())