def test_tts_espeak(self):
config = misc.load_config('.speechrc')
tts = TTS(config.get('tts', 'host'), int(config.get('tts', 'port')))
tts.engine = 'espeak'
first = True
for v, word, ph in ESPEAK_TESTS:
tts.locale = v
tts.voice = v
espeak_ph = tts.gen_ipa(word)
self.assertEqual(espeak_ph, ph)
wav = tts.synthesize(word)
logging.debug('wav len: %d bytes.' % len(wav))
self.assertGreater(len(wav), 100)
wav = tts.synthesize(ph, mode='ipa')
logging.debug('wav len: %d bytes.' % len(wav))
self.assertGreater(len(wav), 100)
if first:
tts.say(word)
first = False
def test_tts_mary(self):
config = misc.load_config('.speechrc')
tts = TTS(config.get('tts', 'host'), int(config.get('tts', 'port')))
# test mary
tts.engine = 'mary'
for l, voice, word, ph in MARY_TESTS:
tts.locale = l
tts.voice = voice
mary_ph = tts.gen_ipa(word)
self.assertEqual(mary_ph, ph)
wav = tts.synthesize(word)
logging.debug('wav len: %d bytes.' % len(wav))
self.assertGreater(len(wav), 100)
wav = tts.synthesize(ph, mode='ipa')
logging.debug('wav len: %d bytes.' % len(wav))
self.assertGreater(len(wav), 100)
def __init__(self, lex):
self.lex = lex
#
# TTS (for audio output)
#
self.tts = TTS ('local', 0, locale='de', voice='bits3', engine='espeak')
#
# sequitur interface
#
self.si = SeqIf(SEQUITUR_MODEL)
def test_tts_pico(self):
config = misc.load_config('.speechrc')
tts = TTS(config.get('tts', 'host'), int(config.get('tts', 'port')))
tts.engine = 'pico'
for v, word in PICO_TESTS:
tts.locale = v
tts.voice = v
wav = tts.synthesize(word)
logging.debug('wav len: %d bytes.' % len(wav))
self.assertGreater(len(wav), 100)
tts.say(word)
transcript.append(t)
#
# config
#
config = misc.load_config('.speechrc')
vf_login = config.get("speech", "vf_login")
extrasdir = config.get("speech", "extrasdir_%s" % lang)
#
# TTS (for audio output)
#
tts = TTS('local', 0, locale='de', voice='bits3', engine='espeak')
#
# load lexicon
#
logging.info("loading lexicon...")
lex = Lexicon(file_name=options.lang)
logging.info("loading lexicon...done.")
#
# main ui loop
#
next_segment()
class Intent(Enum):
HELLO = 1
LIGHT = 2
RADIO = 3
print("Initializing...")
radio_on = False
lights_on = False
asr = ASR(model_dir=MODELDIR)
rec = PulseRecorder(volume=VOLUME)
vad = VAD()
tts = TTS(engine="espeak", voice="en")
utt_map = {}
def add_utt(utterance, intent):
utt_map[utterance] = intent
add_utt("hello computer", Intent.HELLO)
add_utt("switch on the lights", Intent.LIGHT)
add_utt("switch off the lights", Intent.LIGHT)
add_utt("switch on the radio", Intent.RADIO)
add_utt("switch off the radio", Intent.RADIO)
rec.start_recording()
class LexEdit(object):
def __init__(self, lex):
self.lex = lex
#
# TTS (for audio output)
#
self.tts = TTS ('local', 0, locale='de', voice='bits3', engine='espeak')
#
# sequitur interface
#
self.si = SeqIf(SEQUITUR_MODEL)
def lex_gen_ipa (self, lex_base, locale, engine, voice, speak=False):
ipas = u''
try:
if engine == 'sequitur':
# ipas = sequitur_gen_ipa (SEQUITUR_MODEL, lex_base)
ipas = self.si.g2p(lex_base)
else:
self.tts.locale = locale
self.tts.engine = engine
self.tts.voice = voice
ipas = self.tts.gen_ipa (lex_base)
if speak:
self.tts.locale = 'de'
self.tts.engine = 'mary'
self.tts.voice = 'dfki-pavoque-neutral-hsmm'
self.tts.say_ipa(ipas, async=True)
except:
logging.error('EXCEPTION CAUGHT %s' % traceback.format_exc())
return ipas
def edit(self, lex_token):
lex_base = lex_token.split('_')[0]
if lex_token in self.lex:
lex_entry = lex[lex_token]
else:
ipas = self.lex_gen_ipa(lex_base, 'de', 'sequitur', 'de')
lex_entry = {'ipa': ipas}
self.lex[lex_token] = lex_entry
ipas = lex_entry['ipa']
lex_gen = {}
lex_gen['de-mary'] = self.lex_gen_ipa(lex_base, 'de', 'mary', 'bits3')
lex_gen['de-espeak'] = self.lex_gen_ipa(lex_base, 'de', 'espeak', 'de')
lex_gen['de-sequitur'] = self.lex_gen_ipa(lex_base, 'de', 'sequitur', 'de')
try:
self.tts.locale ='de'
self.tts.engine ='mary'
self.tts.voice ='dfki-pavoque-neutral-hsmm'
self.tts.say_ipa(ipas, async=True)
except:
logging.error('EXCEPTION CAUGHT %s' % traceback.format_exc())
while True:
print
print u"Token : %s" % lex_token
print u"IPA : %s" % lex_entry['ipa']
print
for engine in sorted(lex_gen):
print u"%-11s : %s" % (engine, lex_gen[engine])
print
if lex_token in self.lex:
m = self.lex.get_multi(lex_token)
for k in m:
print u"%s [%s]" % (k, m[k]['ipa'])
else:
print u"NEW TOKEN"
print u"SPEAK P:de-unitsel O:de-hsmm I:fr-hsmm U:en-hsmm"
print u"GEN G:de-mary H:de-espeak J:de-sequitur K:fr-mary L:en-mary"
print u" E:Edit Q:Quit "
try:
resp = raw_input("Lex> ")
# quit
if resp.lower() == 'q':
break
# generate de-mary
elif resp.lower() == 'g':
lex_entry['ipa'] = self.lex_gen_ipa (lex_base, 'de', 'mary', 'bits3', True)
# generate de-espeak
elif resp.lower() == 'h':
lex_entry['ipa'] = self.lex_gen_ipa (lex_base, 'de', 'espeak', 'de', True)
# generate en-mary
elif resp.lower() == 'l':
self.tts.locale ='en-US'
self.tts.engine ='mary'
self.tts.voice ='cmu-rms-hsmm'
ipas = self.tts.gen_ipa (lex_base)
self.tts.say_ipa(ipas, async=True)
lex_entry['ipa'] = ipas
# generate fr-mary
elif resp.lower() == 'k':
self.tts.locale ='fr'
self.tts.engine ='mary'
self.tts.voice ='upmc-pierre-hsmm'
ipas = self.tts.gen_ipa (lex_base)
self.tts.say_ipa(ipas, async=True)
lex_entry['ipa'] = ipas
# generate de-sequitur
elif resp.lower() == 'j':
lex_entry['ipa'] = self.lex_gen_ipa (lex_base, 'de', 'sequitur', 'de', True)
# speak de mary unitsel
elif resp.lower() == 'p':
if len(lex_entry['ipa']) == 0:
continue
ipas = lex_entry['ipa']
self.tts.locale ='de'
self.tts.engine ='mary'
self.tts.voice ='bits3'
self.tts.say_ipa(ipas, async=True)
# speak de mary hsmm
elif resp.lower() == 'o':
if len(lex_entry['ipa']) == 0:
continue
ipas = lex_entry['ipa']
self.tts.locale = 'de'
self.tts.engine = 'mary'
self.tts.voice = 'dfki-pavoque-neutral-hsmm'
self.tts.say_ipa(ipas, async=True)
# speak fr mary hsmm
elif resp.lower() == 'i':
if len(lex_entry['ipa']) == 0:
continue
ipas = lex_entry['ipa']
self.tts.locale ='fr'
self.tts.engine ='mary'
self.tts.voice ='upmc-pierre-hsmm'
self.tts.say_ipa(ipas, async=True)
# speak en mary hsmm
elif resp.lower() == 'u':
ipas = lex_entry['ipa']
self.tts.locale = 'en-US'
self.tts.engine = 'mary'
self.tts.voice = 'cmu-rms-hsmm'
self.tts.say_ipa(ipas, async=True)
# edit XS
elif resp.lower() == 'e':
ipas = lex_entry['ipa']
xs = ipa2xsampa (lex_token, ipas, stress_to_vowels=False)
readline.add_history(xs)
xs = raw_input(xs + '> ')
ipas = xsampa2ipa (lex_token, xs)
lex_entry['ipa'] = ipas
except:
logging.error('EXCEPTION CAUGHT %s' % traceback.format_exc())
self.lex.save()
print "new lexicon saved."
print
kernal.setup_nlp_model()
ctx = kernal.create_context()
logging.debug('AI kernal initialized.')
#
# ASR
#
asr = ASR(model_dir=options.asr_model)
logging.debug('ASR initialized.')
#
# TTS
#
tts = TTS(engine="espeak", voice="en")
#
# main loop
#
print(chr(27) + "[2J")
while True:
#
# record audio, run VAD
#
print "Please speak.",
rec.start_recording()
#
# config
#
config = misc.load_config('.speechrc')
wav16_dir = config.get("speech", "wav16")
host = config.get('tts', 'host')
port = int(config.get('tts', 'port'))
#
# TTS
#
tts = TTS (host, port, locale='de', voice='bits3', engine='espeak')
def paint_main(stdscr, cur_ts):
global edit_ts, prompt_tokens, prompt_token_idx
ts = edit_ts[cur_ts]
stdscr.clear()
my, mx = stdscr.getmaxyx()
for x in range(mx):
stdscr.insstr( 0, x, ' ', curses.A_REVERSE)
stdscr.insstr(my-2, x, ' ', curses.A_REVERSE)
stdscr.insstr(my-1, x, ' ', curses.A_REVERSE)
asr = ASR(engine=ASR_ENGINE_NNET3,
model_dir=kaldi_model_dir,
model_name=kaldi_model,
kaldi_beam=kaldi_beam,
kaldi_acoustic_scale=kaldi_acoustic_scale,
kaldi_frame_subsampling_factor=kaldi_frame_subsampling_factor)
logging.debug('ASR initialized.')
#
# TTS
#
tts = TTS(host_tts=tts_host,
port_tts=tts_port,
locale=tts_locale,
voice=tts_voice,
engine=tts_engine,
speed=tts_speed,
pitch=tts_pitch)
#
# main loop
#
print(chr(27) + "[2J")
while True:
#
# record audio, run VAD
#
stdscr.keypad(1)
#
# config
#
config = misc.load_config('.speechrc')
host = config.get('tts', 'host')
port = int(config.get('tts', 'port'))
#
# TTS Client
#
tts = TTS(host, port, locale='de', voice='bits3')
#
# main
#
try:
lex_gen = {}
lex_cur_token = 0
lex_set_token(lex_tokens[lex_cur_token])
while True:
lex_paint_main()
stdscr.keypad(1)
#
# config
#
config = misc.load_config('.speechrc')
host = config.get('tts', 'host')
port = int(config.get('tts', 'port'))
#
# TTS Client
#
tts = TTS(host, port, locale='de', voice='bits3')
#
# main
#
try:
lex_gen = {}
lex_cur_token = 0
lex_set_token(lex_tokens[lex_cur_token])
while True:
lex_paint_main()
transcripts = Transcripts(corpus_name=corpus)
print "loading transcripts...done."
#
# config
#
wav16_dir = config.get("speech", "wav16")
host = config.get('tts', 'host')
port = int(config.get('tts', 'port'))
#
# TTS
#
tts = TTS (host, port, locale='de', voice='bits3')
#
# count
#
known = set()
for spk in spk2gender:
known.add(spk)
num_unk = 0
for cfn in transcripts:
ts = transcripts[cfn]
spk = ts['spk']
kernal.setup_nlp_model()
ctx = kernal.create_context()
logging.debug ('AI kernal initialized.')
#
# ASR
#
asr = ASR(model_dir = options.asr_model)
logging.debug ('ASR initialized.')
#
# TTS
#
tts = TTS(engine="espeak", voice="en")
#
# main loop
#
print(chr(27) + "[2J")
while True:
#
# record audio, run VAD
#
print "Please speak.",
rec.start_recording()
kernal.setup_nlp_model()
ctx = kernal.create_context()
logging.debug('AI kernal initialized.')
#
# ASR
#
asr = ASR(model_dir=options.asr_model)
logging.debug('ASR initialized.')
#
# TTS
#
tts = TTS(engine="espeak", voice=kernal.lang)
#
# main loop
#
print(chr(27) + "[2J")
while True:
#
# record audio, run VAD
#
print "Please speak.",
rec.start_recording()
#!/usr/bin/env python3
from nltools.tts import TTS
tts = TTS(engine="espeak", voice="en")
tts.say("hello from your pi")
