def text_to_speech(self,
text: str,
slow: bool = False,
use_cache: bool = True) -> str:
from gtts import gTTS
file_path = self._get_cache_file_path(text=text, use_cache=use_cache)
if self._is_valid_cached_file(file_path=file_path,
use_cache=use_cache):
LOG.debug("Using existing cached file: %s" % (file_path))
return file_path
LOG.trace('Performing TTS on text "%s" and saving result to %s' %
(text, file_path))
lang = "en-US"
# Sometimes API returns "Unable to find token seed" error so we retry up to 3 times
for index in range(0, 3):
try:
audio_file = gTTS(text=text,
lang=lang,
slow=slow,
lang_check=False)
audio_file.save(file_path)
# Even on successful response API sometimes returns a corrupted response so we need
# to retry on such scenario as well
# TODO: Could we just perform file size check to speed the operation up?
get_audio_file_duration(file_path)
break
except ValueError as e:
if "Unable to find token seed" not in str(e):
raise e
LOG.debug(
"Retrying gtts call due to failure: %s (attempt=%s)" %
(str(e), (index + 1)))
except mutagen.mp3.HeaderNotFoundError as e:
LOG.debug(
"Retrying gtts call due to failure: %s (attempt=%s)" %
(str(e), (index + 1)))
if not os.path.isfile(file_path) or os.stat(file_path).st_size == 0:
LOG.error('Failed to perform TTS on text "%s"' % (text))
return ""
return file_path
def test_text_to_speech_gtts(self):
tts = TextToSpeech(implementation="gtts")
text = "Hello World. Test."
output_file_path = tts._tts._get_cache_file_path(text=text,
use_cache=True)
self.assertFalse(os.path.isfile(output_file_path))
# This test is sometimes flaky when running on CI and "Unable to find token seed" error is
# throw. To try to mitigate this, we retry on failure.
for i in range(0, 3):
try:
tts.text_to_speech(text=text)
break
except ValueError as e:
if "Unable to find token seed" not in str(e):
raise e
print("Retrying gtts call due to failure: %s" % (str(e)))
time.sleep(random.randint(2, 5))
self.assertTrue(os.path.isfile(output_file_path))
duration = get_audio_file_duration(file_path=output_file_path)
self.assertTrue(duration >= 1.5)
def duration(self):
if not self._duration:
if self.type in ["text", "text_to_morse"]:
self._duration = CronSayPlugin.calculate_duration_for_text(
text=self.value)
elif self.type == "morse":
# TODO: Use more accurate estimation for more code
self._duration = CronSayPlugin.calculate_duration_for_text(
text=self.value)
elif self.type == "file":
self._duration = get_audio_file_duration(file_path=self.value)
return self._duration
def test_text_to_speech_speak(self):
tts = TextToSpeech(implementation="espeak")
text = "Hello World. Test."
output_file_path = tts._tts._get_cache_file_path(text=text,
use_cache=True)
self.assertFalse(os.path.isfile(output_file_path))
tts.text_to_speech(text=text)
self.assertTrue(os.path.isfile(output_file_path))
duration = get_audio_file_duration(file_path=output_file_path)
self.assertTrue(duration >= 1.5)
def test_text_to_speech_govornik(self):
tts = TextToSpeech(implementation="govornik")
text = "Dober dan. Dobro jutri vsi."
output_file_path = tts._tts._get_cache_file_path(text=text,
use_cache=True)
self.assertFalse(os.path.isfile(output_file_path))
for i in range(0, 3):
try:
tts.text_to_speech(text=text, language="sl_SI")
break
except Exception as e:
print("Retrying govornik call due to failure: %s" % (str(e)))
time.sleep(random.randint(2, 5))
self.assertTrue(os.path.isfile(output_file_path))
duration = get_audio_file_duration(file_path=output_file_path)
self.assertTrue(duration >= 1.5)
def test_get_audio_file_duration(self):
duration = get_audio_file_duration(MP3_FILE_PATH)
self.assertEqual(duration, 6.048)
duration = get_audio_file_duration(WAV_FILE_PATH)
self.assertEqual(duration, 6.048)