def make_espeak(text, lang, max_length):
voice = Voice(lang=gtts_to_espeak[lang], speed=130,
volume=2) if lang in gtts_to_espeak else Voice(lang="en",
speed=130)
wav = voice.to_audio(text)
pydub_wav = AudioSegment.from_file_using_temporary_files(BytesIO(wav))
audio_length = len(pydub_wav) / 1000
return wav, audio_length
class TestEffects(unittest.TestCase):
text = "Les écoute pas ces sheitane c moi le vrai crysw"
def setUp(self):
self.effects = {
eff_class: set()
for eff_class in (PhonemicEffect, TextEffect, AudioEffect,
VoiceEffect)
}
for effect_cls in set(AVAILABLE_EFFECTS):
effect = effect_cls()
for cls in self.effects.keys():
if isinstance(effect, cls):
self.effects[cls].add(effect)
self.voice = Voice(lang="fr")
def test_text_effects(self):
for effect in self.effects[TextEffect]:
self.assertIsNotNone(effect.process(self.text))
def test_phonemic_effects(self):
pho = self.voice.to_phonemes(self.text)
for effect in self.effects[PhonemicEffect]:
self.assertIsNotNone(effect.process(pho))
def test_audio_effects(self):
wav = self.voice.to_audio(self.text)
_, wav_array = get_event_loop().run_until_complete(
AudioRenderer.to_f32_16k(wav))
for effect in self.effects[AudioEffect]:
self.assertIsNotNone(effect.process(wav_array))
def test_voice_effects(self):
cookie_hash = md5(("parce que nous we").encode('utf8')).digest()
voice_params = VoiceParameters.from_cookie_hash(cookie_hash)
for effect in self.effects[VoiceEffect]:
self.assertIsNotNone(effect.process(voice_params))
beat_duration = len(beats_track) / (rate * BEATS_TRACK_BEAT_COUNT)
beats_track_looped = np.tile(
beats_track,
LOOPS_COUNT * BEATS_PER_MEASURE * len(CHORDS_PROGRESSION) //
BEATS_TRACK_BEAT_COUNT)
logging.info("Beat time : %dms" % (beat_duration * 1000))
logging.info("Measure time : %dms" % (beat_duration * 1000 * 4))
prog_freqs = get_trinote_progression_freqs(CHORDS_PROGRESSION)
logging.info("First freq progression: \n %s" % (str(prog_freqs)))
track_freqs = prog_freqs * LOOPS_COUNT
progression_phonems = PhonemeList([])
for freq in track_freqs:
progression_phonems.append(
Phoneme("a", int(beat_duration * 1000), [(0, freq), (100, freq)]))
logging.info("Rendering audio")
voice = Voice(lang="fr", voice_id=2)
wav = voice.to_audio(progression_phonems)
if BEATS_TRACK_FILE is not None:
rate, wave_array = to_f32_16k(wav)
mixed_tracks = mix_tracks(beats_track_looped * 0.6,
wave_array * 0.3,
align="left")
wav = to_wav_bytes(mixed_tracks, 16000)
player = AudioPlayer()
player.set_file(BytesIO(wav))
player.play()
player.close()
beats_track_looped = np.tile(beats_track,
PROG_LOOP_COUNT * BEATS_PER_MEASURE * FULL_LOOP_COUNT * \
(FIRST_PROG_REPEAT + SECOND_PROG_REPEAT) // BEATS_TRACK_BEAT_COUNT)
logging.info("Beat time : %dms" % (beat_duration * 1000))
logging.info("Measure time : %dms" % (beat_duration * 1000 * 4))
first_prog_freqs = get_trinote_progression_freqs(FIRST_CHORDS_PROGRESSION)
logging.info("First freq progression: \n %s" % (str(first_prog_freqs)))
second_prog_freqs = get_trinote_progression_freqs(SECOND_CHORDS_PROGRESSION)
logging.info("First freq progression: \n %s" % (str(second_prog_freqs)))
track_freqs = first_prog_freqs * FIRST_PROG_REPEAT + second_prog_freqs * SECOND_PROG_REPEAT
progression_phonems = PhonemeList([])
for freq in track_freqs * FULL_LOOP_COUNT:
progression_phonems.append(
Phoneme("w", int(beat_duration * 1000), [(0,freq), (100,freq)])
)
logging.info("Rendering audio")
voice = Voice(lang="fr", voice_id=3)
wav = voice.to_audio(progression_phonems)
if BEATS_TRACK_FILE is not None:
rate, wave_array = to_f32_16k(wav)
mixed_tracks = mix_tracks(beats_track_looped * 0.6, wave_array * 0.4, align="left")
wav = to_wav_bytes(mixed_tracks, 16000)
player = AudioPlayer()
player.set_file(BytesIO(wav))
player.play()
player.close()