async def get_audio_response(self):
"""Download audio data then send to speech-to-text API for answer"""
try:
audio_url = await self.image_frame.evaluate(
'jQuery("#audio-source").attr("src")')
if not isinstance(audio_url, str):
raise DownloadError(f"Audio url is not valid, expected `str`"
"instead received {type(audio_url)}")
except CancelledError:
raise DownloadError("Audio url not found, aborting")
self.log("Downloading audio file")
try:
audio_data = await self.loop.create_task(
util.get_page(audio_url,
proxy=self.proxy,
proxy_auth=self.proxy_auth,
binary=True,
timeout=self.page_load_timeout))
except ClientError as e:
self.log(f"Error `{e}` occured during audio download, retrying")
else:
answer = None
service = self.speech_service.lower()
if service in [
"azure", "pocketsphinx", "deepspeech", "azurespeech"
]:
if service == "azurespeech":
speech = AzureSpeech()
elif service == "azure":
speech = Azure()
elif service == "pocketsphinx":
speech = Sphinx()
else:
speech = DeepSpeech()
tmpd = tempfile.mkdtemp()
tmpf = os.path.join(tmpd, "audio.mp3")
await util.save_file(tmpf, data=audio_data, binary=True)
answer = await self.loop.create_task(speech.get_text(tmpf))
shutil.rmtree(tmpd)
else:
speech = Amazon()
answer = await self.loop.create_task(
speech.get_text(audio_data))
if answer:
self.log(f'Received answer "{answer}"')
return answer
self.log("No answer, reloading")
await self.click_reload_button()
func = (
f'"{audio_url}" !== '
f'jQuery(".rc-audiochallenge-tdownload-link").attr("href")')
try:
await self.image_frame.waitForFunction(
func, timeout=self.animation_timeout)
except TimeoutError:
raise ReloadError("Download link never updated")
async def solve(service, audio_file):
answer = None
service = service.lower()
if service in ["azure", "pocketsphinx", "deepspeech"]:
if service == "azure":
speech = Azure()
elif service == "pocketsphinx":
speech = Sphinx()
else:
speech = DeepSpeech()
answer = await speech.get_text(audio_file)
else:
speech = Amazon()
answer = await speech.get_text(audio_file)
if answer:
print(f"audio file: {audio_file}")
print(f"{service}'s best guess: {answer}")
else:
print(f"{service}'s failed to decipher: {audio_file}")
if post_play_audio:
await play_audio(audio_file)
print("--------" * 15)