def _jtalk_speak(msg, index=None, prop=None):
global currIndex, buff
global currentEngine
global lastIndex
if prop is None: return
currIndex = index
if prop.characterMode:
fperiod_current = voice_args['fperiod']
else:
fperiod_current = fperiod
msg = unicode_normalize(msg)
msg = jtalkPrepare.convert(msg)
lw = None
if DEBUG: lw = logwrite
setSpeaking(True)
currentEngine = 2
if DEBUG: logwrite("p:%d i:%d msg:%s" % (prop.pitch, prop.inflection, msg))
level = int(max_level * speaker_attenuation)
la = 0.020 * (prop.inflection + voice_args.get('inflection_bias', 0)) # 50 = original range
ls = 0.015 * (prop.pitch - 50.0 + voice_args['pitch_bias']) # 50 = no shift
lo = ls + voice_args['lf0_base'] * (1 - la)
if DEBUG: logwrite("lo:%f la:%f" % (lo, la))
lastIndex = currIndex
currIndex = None
for t in string.split(msg):
if DEBUG: logwrite("unicode (%s)" % t)
s = text2mecab(t)
if DEBUG: logwrite("utf-8 (%s)" % s.decode('utf-8', 'ignore'))
if not isSpeaking(): libjt_refresh(); return
mf = MecabFeatures()
Mecab_analysis(s, mf, logwrite_=logwrite)
if DEBUG: Mecab_print(mf, logwrite)
Mecab_correctFeatures(mf)
if DEBUG: Mecab_print(mf, logwrite)
ar = Mecab_splitFeatures(mf, CODE_='utf-8')
for a in ar:
if isSpeaking():
if DEBUG: Mecab_print(a, logwrite, CODE_='utf-8')
Mecab_utf8_to_cp932(a)
if DEBUG: logwrite("Mecab_analysis done")
libjt_synthesis(
a.feature,
a.size,
fperiod_ = fperiod_current,
feed_func_ = player.feed, # player.feed() is called inside
is_speaking_func_ = isSpeaking,
begin_thres_ = thres_level,
end_thres_ = thres2_level,
level_ = level,
logwrite_ = lw,
lf0_offset_ = lo,
lf0_amp_ = la)
libjt_refresh()
if DEBUG: logwrite("libjt_synthesis done")
del a
del mf
player.idle()
setSpeaking(False)
currentEngine = 0
def runTasks():
jtalkPrepare.setup()
count = 0
for item in tests:
msg = item[0]
msg = unicode_normalize(msg)
s = jtalkPrepare.convert(msg)
if item[1] != s:
_print("expected:%s result:%s" % (item[1], s))
count += 1
return count
def runTasks():
jtalkPrepare.setup()
count = 0
for item in tests:
msg = item[0]
normalized = unicode_normalize(msg)
s = jtalkPrepare.convert(normalized)
if item[1] != s:
_print('input:%s normalized:%s result:%s expected:%s' % (msg, normalized, s, item[1]))
count += 1
return count
def do_synthesis(msg, voice_args, do_play, do_write, do_log, fperiod, pitch=50, inflection=50, vol=50):
global count
msg = jtalkPrepare.convert(msg)
s = text2mecab(msg)
__print("utf-8: (%s)" % s.decode('utf-8', 'ignore'))
mf = MecabFeatures()
Mecab_analysis(s, mf)
Mecab_print(mf, __print)
Mecab_correctFeatures(mf)
ar = [mf] #ar = Mecab_splitFeatures(mf)
__print('array size %d' % len(ar))
max_level = int(326.67 * int(vol) + 100) # 100..32767
level = int(max_level * voice_args['speaker_attenuation'])
lf0_amp = 0.020 * inflection # 50 = original range
ls = 0.015 * (pitch - 50.0 + voice_args['pitch_bias']) # 50 = no shift
lf0_offset = ls + voice_args['lf0_base'] * (1 - lf0_amp)
for a in ar:
count += 1
__print('feature size %d' % a.size)
Mecab_print(a, __print)
Mecab_utf8_to_cp932(a)
if do_write:
w = "_test%d.jt.wav" % count
else:
w = None
if do_log:
l = "_test%d.jtlog" % count
else:
l = None
data = libjt_synthesis(a.feature,
a.size,
begin_thres_=32,
end_thres_=32,
level_=level,
fperiod_ = fperiod,
lf0_offset_ = lf0_offset,
lf0_amp_ = lf0_amp,
logwrite_ = __print,
jtlogfile_ = l,
jtwavfile_ = w)
if data:
__print('data size %d' % len(data))
if do_play:
pa_play(data, samp_rate = voice_args['samp_rate'])
if do_write:
w = wave.Wave_write("_test%d.wav" % count)
w.setparams( (1, 2, voice_args['samp_rate'], len(data)/2,
'NONE', 'not compressed') )
w.writeframes(data)
w.close()
libjt_refresh()
del a
del mf
def do_synthesis(msg, voice_args, do_play, do_write, do_write_jt, do_log, fperiod, pitch=50, inflection=50, vol=50):
msg = jtalkPrepare.convert(msg)
s = text2mecab(msg)
__print("utf-8: (%s)" % s.decode('utf-8', 'ignore'))
mf = MecabFeatures()
Mecab_analysis(s, mf)
Mecab_print(mf, __print)
Mecab_correctFeatures(mf)
ar = Mecab_splitFeatures(mf)
__print('array size %d' % len(ar))
max_level = int(326.67 * int(vol) + 100) # 100..32767
level = int(max_level * voice_args['speaker_attenuation'])
lf0_amp = 0.020 * inflection # 50 = original range
ls = 0.015 * (pitch - 50.0 + voice_args['pitch_bias']) # 50 = no shift
lf0_offset = ls + voice_args['lf0_base'] * (1 - lf0_amp)
count = 0
for a in ar:
count += 1
__print('feature size %d' % a.size)
Mecab_print(a, __print)
Mecab_utf8_to_cp932(a)
if do_write_jt:
w = "_test%d.jt.wav" % count
else:
w = None
if do_log:
l = "_test%d.jtlog" % count
else:
l = None
data = libjt_synthesis(a.feature,
a.size,
begin_thres_=32,
end_thres_=32,
level_=level,
fperiod_ = fperiod,
lf0_offset_ = lf0_offset,
lf0_amp_ = lf0_amp,
logwrite_ = __print,
jtlogfile_ = l,
jtwavfile_ = w)
if data:
__print('data size %d' % len(data))
if do_play:
pa_play(data, samp_rate = voice_args['samp_rate'])
if do_write:
w = wave.Wave_write("_test%d.wav" % count)
w.setparams( (1, 2, voice_args['samp_rate'], len(data)/2,
'NONE', 'not compressed') )
w.writeframes(data)
w.close()
libjt_refresh()
del a
del mf