def main():
# See http://g.co/cloud/speech/docs/languages
# for a list of supported languages.
language_code = 'en-US' # a BCP-47 language tag
# If enabled, each word in the first alternative of each result will be
# tagged with a speaker tag to identify the speaker.
enable_speaker_diarization = True
# Optional. Specifies the estimated number of speakers in the conversation.
# diarization_speaker_count = 2
client = speech_v1p1beta1.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=RATE,
language_code=language_code,
enable_speaker_diarization=enable_speaker_diarization)
streaming_config = types.StreamingRecognitionConfig(config=config,
interim_results=True)
with MicrophoneStream(RATE, CHUNK) as stream:
audio_generator = stream.generator()
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = client.streaming_recognize(streaming_config, requests)
# Now, put the transcription responses to use.
listen_print_loop(responses)
def google_transcribe_file(fp, bucket_name='prof-resp-trans'):
storage_client = google.cloud.storage.Client()
bucket = storage_client.get_bucket(bucket_name)
client = google.cloud.speech_v1p1beta1.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.ENCODING_UNSPECIFIED,
language_code='en-US',
enable_word_time_offsets=True,
enable_automatic_punctuation=True)
blob = bucket.blob(fp.name)
if not blob.exists():
new_print('Uploading File: {}'.format(fp.name))
blob.upload_from_filename(str(fp.resolve()))
new_print('Finished Uploading: {}'.format(fp.name))
else:
new_print('File already uploaded: {}'.format(fp.name))
new_print('Starting transcription...')
audio = types.RecognitionAudio(
uri='gs://{}/{}'.format(bucket_name, fp.name))
response = client.long_running_recognize(config, audio)
results = response.result()
new_print('Transciption finished')
return results
def read(file_name, result_file, time=50):
# Instantiates a client
client = speech.SpeechClient()
# The name of the audio file to transcribe
file_name = os.path.join(
os.path.dirname(__file__),
file_name)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
audio_channel_count=2,
sample_rate_hertz=44100,
enable_speaker_diarization=True,
diarization_speaker_count=2,
language_code='ko-KR')
# Loads the audio into memory
with io.open(file_name, 'rb') as audio_file:
with io.open(result_file, 'w') as f:
FILE_BYTE = 44100 * 2
read_byte = FILE_BYTE * time
while True:
content = audio_file.read(read_byte)
if not content:
break
audio = types.RecognitionAudio(content=content)
response = client.recognize(config, audio)
for result in response.results:
f.write(result.alternatives[0].transcript)
print(result.alternatives[0])
print('Transcript: {}'.format(result.alternatives[0].transcript))
def configureAPI(self):
if self.title.find('.flac') != -1:
with open(os.path.join(self.path, self.title), 'rb') as audio_file:
content = audio_file.read()
self.audio = types.RecognitionAudio(content=content)
else:
self.audio = types.RecognitionAudio(uri="gs://twiml-mp3/" +
self.title + ".flac")
self.config = types.RecognitionConfig(
encoding=self.encoding,
sample_rate_hertz=self.sample_rate,
language_code=self.language_code,
enable_automatic_punctuation=self.punctuation,
enable_speaker_diarization=self.diarization,
diarization_speaker_count=self.num_speakers,
audio_channel_count=1,
use_enhanced=self.enhanced,
model=self.model,
enable_word_time_offsets=self.time_offsets,
enable_word_confidence=self.word_confidence,
max_alternatives=self.max_alternatives,
metadata=self.metadata,
speech_contexts=[types.SpeechContext(phrases=self.phrases)])
def transcribe_gcs(gcs_uri, hint_phrases, set_config):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
client = speech_v1p1beta1.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
# hint_phrase = []
# set_config['enable_speaker_diarization'] = 'False'
print(set_config.get('enable_automatic_punctuation'))
# Set default values, check dict having each key and cast from str to each type.
config = types.RecognitionConfig(
encoding=eval(set_config.get('encoding', 'enums.RecognitionConfig.AudioEncoding.FLAC')),
sample_rate_hertz=int(set_config.get('sample_rate_hertz', 16000)),
language_code=set_config.get('language_code', 'en-US'),
enable_automatic_punctuation=eval(set_config.get('enable_automatic_punctuation', True)),
enable_speaker_diarization=eval(set_config.get('enable_speaker_diarization', False)),
diarization_speaker_count=int(set_config.get('diarization_speaker_count', 1)),
speech_contexts=[speech_v1p1beta1.types.SpeechContext(phrases=hint_phrases)])
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
response = operation.result(timeout=900)
return response
def transcribe_audio(self, audio_blob):
phrases = [c.command_variant for c in self.commands]
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
language_code='en-US',
audio_channel_count=audio_blob.pop('n_channels'),
enable_word_time_offsets=True,
model='video',
speech_contexts=[
dict(phrases=phrases, boost=self.commandword_bias)
])
try:
operation = self.client.long_running_recognize(config, audio_blob)
except ResourceExhausted:
err_msg = f"The project has run out of it's quota for today. Try again tomorrow or set up your own Google Cloud project, see '{meta_utils.install_url()}'"
print(err_msg)
sys.exit(1)
print(u"Analyzing speech...")
response = operation.result()
words = []
for result in response.results:
for word in result.alternatives[0].words:
words.append(word)
return words
def voice_recognize(storage_uri):
"""
Performs synchronous speech recognition on an audio file
Args:
storage_uri URI for audio file in Cloud Storage, e.g. gs://[BUCKET]/[FILE]
"""
client = speech_v1p1beta1.SpeechClient()
# storage_uri = 'gs://cloud-samples-data/speech/brooklyn_bridge.mp3'
# Encoding of audio data sent. This sample sets this explicitly.
# This field is optional for FLAC and WAV audio formats.
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.MP3,
sample_rate_hertz=44100,
language_code='en-US',
# Enable automatic punctuation
enable_automatic_punctuation=True)
audio = {"uri": storage_uri}
response = client.recognize(config, audio)
result = [r.alternatives[0].transcript for r in response.results]
return ' '.join(result)
def microphone_streaming_start(wf, output_stream):
# See http://g.co/cloud/speech/docs/languages
# for a list of supported languages.
language_code = 'ko-KR'
client = speech.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=RATE,
language_code=language_code,
enable_automatic_punctuation=True,
enable_word_time_offsets=True)
# enable_speaker_diarization=True,
# diarization_speaker_count=3)
streaming_config = types.StreamingRecognitionConfig(config=config,
interim_results=True)
with MicrophoneStream(RATE, CHUNK, wf, output_stream) as stream:
audio_generator = stream.generator()
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = client.streaming_recognize(streaming_config, requests)
# Now, put the transcription responses to use.
listen_print_loop(responses)
def transcribe_file_with_diarization():
audio = types.RecognitionAudio(uri=args['input'])
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
# encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
sample_rate_hertz=22050,
language_code='en-US',
enable_speaker_diarization=True,
diarization_speaker_count=2,
model='phone_call')
operation = client.long_running_recognize(config, audio)
print("Waiting on response from google cloud...")
response = operation.result(timeout=720) ## 360 call 01
for result in response.results:
print("\n\n::BEGIN TRANSCRIPT::\n")
print("{}".format(result.alternatives[0].transcript))
print("\n::END TRANSCRIPT::\n\n")
print("\t\tCONFIDENCE: {} \n\n".format(
result.alternatives[0].confidence))
print("::BEGIN SPEAKER DIARIZATION::\n")
words_info = result.alternatives[0].words
for word_info in words_info:
print("{}: '{}'".format(word_info.speaker_tag, word_info.word))
print("\n::END SPEAKER DIARIZATION")
def convertWAVToTranscript(fileFullPathname, split_length_inSeconds):
pool = Pool(8) # Number of concurrent threads
#spk = new SpeechContext { phrases : [ "લગભગ", "માત્ર","શ્રી ત્રંબકભાઈ","સોભાગભાઈ","દેહ વિલય","જ્ઞાની પુરુષ","દશા","રુવાડા","ઐશ્વર્ય","એ","જ્ઞાન","વિકલ્પ","ત્યારે","હમણાં","મુમુક્ષુ","દશા","માર્ગ","અદ્દભુત","નિશ્ચય","સ્મૃતિ" ] };
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
#sample_rate_hertz=512000,
language_code='gu-IN',
speechContexts = {"phrases" : ['લગભગ', 'માત્ર','શ્રી ત્રંબકભાઈ','સોભાગભાઈ','દેહ વિલય','જ્ઞાની પુરુષ','દશા','રુવાડા','ઐશ્વર્ય','એ','જ્ઞાન','વિકલ્પ','ત્યારે','હમણાં']} #,'મુમુક્ષુ','દશા','માર્ગ','અદ્દભુત','નિશ્ચય','સ્મૃતિ'])]
#,"મુમુક્ષુ","દશા","માર્ગ","અદ્દભુત","નિશ્ચય","સ્મૃતિ" ])]
)
files = [f for f in listdir(fileFullPathname) if isfile(join(fileFullPathname, f))]
#file_Direcotry = os.path.dirname(fileFullPathname)
def transcribe(data):
idx, file = data
num, _ = file.split('.')
text_script = ""
name = file
#print(file + " - started")
# Loads the audio into memory
with io.open(fileFullPathname+'\\'+file, 'rb') as audio_file:
content = audio_file.read()
audio = types.RecognitionAudio(content=content)
# Transcribe audio file
# Detects speech in the audio file
client = speech.SpeechClient()
response = client.recognize(config, audio)
for result in response.results:
text_script += result.alternatives[0].transcript
#print(name + " - done")
return {
"idx": num,
"text": text_script
}
all_text = pool.map(transcribe, enumerate(files))
pool.close()
pool.join()
transcript = ""
total_seconds = 0
for t in sorted_nicely(all_text): #sorted(all_text, key=lambda x: sorted_nicely(x['idx'])):
#total_seconds += split_length_inSeconds
print("Duration of file {} is {}".format(fileFullPathname+t['idx']+'.wav', math.ceil(get_duration(fileFullPathname+'\\'+t['idx']+'.wav'))))
total_seconds += math.ceil(get_duration(fileFullPathname+t['idx']+'.wav'))
# Cool shortcut from:
# https://stackoverflow.com/questions/775049/python-time-seconds-to-hms
# to get hours, minutes and seconds
m, s = divmod(total_seconds, 60)
h, m = divmod(m, 60)
# Format time as h:m:s - 30 seconds of text
transcript = transcript + "{:0>2d}:{:0>2d}:{:0>2d} {}\n".format(h, m, s, t['text'])
#print(transcript)
with open("transcript.txt", "w", encoding='utf-8') as f:
f.write(transcript)
def transcribe_gcs(gcs_uri, hertz, channel):
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.FLAC, # flacの設定
sample_rate_hertz=int(hertz), # ヘルツは音声ファイルに合わせる
audio_channel_count=int(channel),
language_code='ja-JP', # 日本語音声の場合
enable_speaker_diarization=True, # 異なる話者の分離
enable_automatic_punctuation=True, # 句読点
speech_contexts=SELECTED_PHRASES # 音声適応ブースト
)
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
operationResult = operation.result()
filename = gcs_uri.rsplit('/', 1)[1].split('.')[0] + ".txt"
outputfilepath = os.path.join(OUTPUT_FOLDER, filename)
fout = codecs.open(outputfilepath, 'a', 'utf-8')
for result in operationResult.results:
for alternative in result.alternatives:
fout.write(u'{}\n'.format(alternative.transcript))
fout.close()
def diarized_transcribe(gcred, gcs_uri, speakercount):
os.environ['GOOGLE_APPLICATION_CREDENTIALS'] = gcred
client = speech_v1p1beta1.SpeechClient()
audio = beta_types.RecognitionAudio(uri=gcs_uri)
config = beta_types.RecognitionConfig(
encoding=beta_enums.RecognitionConfig.AudioEncoding.FLAC,
language_code='en-US',
enable_speaker_diarization=True,
diarization_speaker_count=speakercount,
enable_word_time_offsets=True,
model='video',
enable_automatic_punctuation=True)
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=3600)
transcript = MessageToDict(response)
transcript = transcript.get('results')
transcript = transcript.pop()
transcript = transcript.get('alternatives')
transcript = transcript.pop()
transcript = transcript.get('words')
return transcript
def run(self):
"""Called from [start]. Connects to service and begins streaming."""
# Exit if stop event occurred.
if self._stop_event.is_set():
return
# Create SSL channel.
channel = self._create_channel()
self.is_started = True
# Open stream
service = cloud_speech.SpeechClient(channel)
streaming_config = types.StreamingRecognitionConfig(
config=types.RecognitionConfig(
enable_automatic_punctuation=self.punctuation,
encoding=self.encoding,
sample_rate_hertz=self.rate,
language_code=self.language,),
interim_results=self.interim_results)
try:
request_stream = self._request_stream()
resp_stream = service.streaming_recognize(
streaming_config, request_stream)
self._handle_results(resp_stream)
finally:
self.stop()
def get_client(lang='en-US',
sample_rate=16000,
interim_results=False,
single_utterance=True,
phrase_key=""):
"""
Helper to return client and config
"""
client = SpeechClient()
config = types.StreamingRecognitionConfig(
config=types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=sample_rate,
language_code=lang,
# Enhanced models are only available to projects that
# opt in for audio data collection.
use_enhanced=True,
# A model must be specified to use enhanced model.
model="command_and_search",
speech_contexts=[
types.SpeechContext(phrases=PhraseGenerator.get_phrases(
"app/config.json", phrase_key), )
]),
interim_results=interim_results,
single_utterance=single_utterance)
print(str(config))
return client, config
def transcribe_gcs(gcs_uri):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
from google.cloud import speech
from google.cloud.speech import enums
from google.cloud.speech import types
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=8000,
language_code='en-US',
profanity_filter=True,
speech_contexts=[
speech.types.SpeechContext(
phrases=['Andy', 'Wisy', 'EEP', 'Project', 'Tom', 'Jeff'], )
],
)
operation = client.long_running_recognize(config, audio)
print('GCS -- Waiting for operation to complete...')
response = operation.result(timeout=90)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
print(('Transcript: {}'.format(result.alternatives[0].transcript)))
print(('Confidence: {}'.format(result.alternatives[0].confidence)))
def audio_main():
f = open(u"Nao_log.txt", u"a")
f.write(
u'##**************************** Audio Log File (Group 1) *********************************##'
)
f.close()
# See http://g.co/cloud/speech/docs/languages
# for a list of supported languages.
language_code = u'en-US' # a BCP-47 language tag
# If enabled, each word in the first alternative of each result will be
# tagged with a speaker tag to identify the speaker.
enable_speaker_diarization = True
# Optional. Specifies the estimated number of speakers in the conversation.
#diarization_speaker_count = 2
client = speech_v1p1beta1.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=RATE,
language_code=language_code,
enable_speaker_diarization=enable_speaker_diarization)
streaming_config = types.StreamingRecognitionConfig(config=config,
interim_results=True)
with MicrophoneStream(RATE, CHUNK) as stream:
while not stream.closed:
sys.stdout.write(YELLOW)
sys.stdout.write(u'\n' + unicode(STREAMING_LIMIT *
stream.restart_counter) +
u': NEW REQUEST\n')
stream.audio_input = []
audio_generator = stream.generator()
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = client.streaming_recognize(streaming_config, requests)
# Now, put the transcription responses to use.
listen_print_loop(responses, stream)
if stream.result_end_time > 0:
stream.final_request_end_time = stream.is_final_end_time
stream.result_end_time = 0
stream.last_audio_input = []
stream.last_audio_input = stream.audio_input
stream.audio_input = []
stream.restart_counter = stream.restart_counter + 1
if not stream.last_transcript_was_final:
sys.stdout.write(u'\n')
stream.new_stream = True
def google_transcribe(audio_file_name):
file_name = filepath + audio_file_name
frame_rate, channels = frame_rate_channel(file_name)
if channels > 1:
stereo_to_mono(file_name)
bucket_name = bucketname
source_file_name = filepath + audio_file_name
destination_blob_name = audio_file_name
upload_blob(bucket_name, source_file_name, destination_blob_name)
gcs_uri = 'gs://' + bucketname + '/' + audio_file_name
transcript = ''
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
print("Setting up configurations")
speech_context = speech.types.SpeechContext(phrases=[
"$OOV_CLASS_DIGIT_SEQUENCE", "$YEAR", "$PERCENT", "$MONEY", "$MONTH"
])
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code='en-US',
enable_speaker_diarization=True,
diarization_speaker_count=2,
speech_contexts=[speech_context],
use_enhanced=True,
model="phone_call")
# Detects speech in the audio file
print("detecting speech")
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
tag = 1
speaker = ""
print("Assembling words")
for word_info in words_info:
if word_info.speaker_tag == tag:
speaker = speaker + " " + word_info.word
else: #Changed
transcript += "speaker {}: {}".format(tag, speaker) + '\n'
tag = word_info.speaker_tag
speaker = "" + word_info.word
transcript += "speaker {}: {}".format(tag, speaker)
delete_blob(bucket_name, destination_blob_name)
return transcript
def sub_main(profanityFilterBool):
"""
*** Code taken from Google Cloud Speech to text documentation ***
Turns on the profanity filter so bad words are censored and not printed
"""
# See http://g.co/cloud/speech/docs/languages
# for a list of supported languages.
language_code = 'en-US' # a BCP-47 language tag
sp_c_cico = {
"phrases": ["Hey cico", "Hey Kiko"],
"boost": 30.0
} # speech_contexts_cico
sp_c_kiko = {
"phrases": ["cico", "Cico", "kiko", "Kiko", "kygo", "Kitty, girl"],
"boost": 0
}
movement_words = {
"phrases" : ["move", "feet", "forward", "right", "left", "backward", "degrees", "radians", "to the left", "to the right"],
"boost": 20.0
}
numbers = {
"phrases": ["one", "two", "three", "four", "five", "six", "seven", "eight", "nine", "ten"],
"boost": 5.0
}
relevant_words = {
"phrases": ["cornell cup robotics", "and", "pick up", "grab"],
"boost": 10.0
}
speech_contexts = [sp_c_cico, sp_c_kiko, movement_words, relevant_words]
client = speech_v1p1beta1.SpeechClient()
# print(help(types.RecognitionConfig))
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=RATE,
language_code=language_code,
enable_automatic_punctuation=True,
speech_contexts=speech_contexts)
streaming_config = types.StreamingRecognitionConfig(
config=config,
interim_results=True)
with MicrophoneStream(RATE, CHUNK) as stream:
audio_generator = stream.generator()
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = client.streaming_recognize(streaming_config, requests)
# Now, put the transcription responses to use.
solution = returnResponseString(responses) # solution is the result
append_to_file("log.txt", str(solution))
return solution
def convertFLACToTranscript(fileFullPathname, split_length_inSeconds):
pool = Pool(16) # Number of concurrent threads
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
#sample_rate_hertz=512000,
language_code='gu-IN')
files = [f for f in listdir(fileFullPathname) if isfile(join(fileFullPathname, f))]
#file_Direcotry = os.path.dirname(fileFullPathname)
def transcribe(data):
idx, file = data
num, _ = file.split('.')
text_script = ""
name = file
print(file + " - started")
# Loads the audio into memory
with io.open(fileFullPathname+'\\'+file, 'rb') as audio_file:
content = audio_file.read()
audio = types.RecognitionAudio(content=content)
# Transcribe audio file
# Detects speech in the audio file
client = speech.SpeechClient()
response = client.recognize(config, audio)
for result in response.results:
text_script += result.alternatives[0].transcript
print(name + " - done")
return {
"idx": num,
"text": text_script
}
all_text = pool.map(transcribe, enumerate(files))
pool.close()
pool.join()
transcript = ""
total_seconds = 0
for t in sorted_nicely(all_text): #sorted(all_text, key=lambda x: sorted_nicely(x['idx'])):
#print("Duration of file {} is {}".format(fileFullPathname+t['idx']+'.flac', math.ceil(get_duration(fileFullPathname+'\\'+t['idx']+'.flac'))))
total_seconds += math.ceil(get_duration(fileFullPathname+t['idx']+'.flac'))
# Cool shortcut from:
# https://stackoverflow.com/questions/775049/python-time-seconds-to-hms
# to get hours, minutes and seconds
m, s = divmod(total_seconds, 60)
h, m = divmod(m, 60)
# Format time as h:m:s - 30 seconds of text
transcript = transcript + "{:0>2d}:{:0>2d}:{:0>2d} {}\n".format(h, m, s, t['text'])
#print(transcript)
with open("transcript.txt", "w", encoding='utf-8') as f:
f.write(transcript)
def transcribe_gcs(gcs_uri):
from google.cloud import speech_v1p1beta1 as speech
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
sample_rate_hertz=32000,
encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
enable_speaker_diarization=True,
diarization_speaker_count=2,
language_code='ja-JP')
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
operationResult = operation.result()
d = datetime.datetime.today()
today = d.strftime("%Y%m%d-%H%M%S")
fout = codecs.open('output{}.txt'.format(today), 'a', 'shift_jis')
speaker_1_words = ""
speaker_1_s = float(0)
speaker_2_words = ""
speaker_2_s = float(0)
for word in operationResult.results[-1].alternatives[0].words:
tmp_word = u'{}'.format(word.word.split("|")[0])
start_time = float(word.start_time.seconds) + float(
word.start_time.nanos) / 1000 / 1000 / 1000
end_time = float(word.end_time.seconds) + float(
word.end_time.nanos) / 1000 / 1000 / 1000
s = end_time - start_time
if word.speaker_tag == 1:
speaker_1_s += s
speaker_1_words += tmp_word
else:
speaker_2_s += s
speaker_2_words += tmp_word
fout.write('speaker_1: \n{}\n'.format(speaker_1_words))
fout.write('s: {}\n'.format(speaker_1_s))
fout.write('speaker_2: \n{}\n'.format(speaker_2_words))
fout.write('s: {}\n'.format(speaker_2_s))
#for result in operationResult.results:
# for alternative in result.alternatives:
# fout.write(u'{}\n'.format(alternative.transcript))
fout.close()
def google_transcribe(audio_file_name):
file_name = filepath + audio_file_name
m4a_to_wav(file_name)
# The name of the audio file to transcribe
frame_rate, channels = frame_rate_channel(file_name)
if channels > 1:
stereo_to_mono(file_name)
bucket_name = bucketname
source_file_name = filepath + audio_file_name
destination_blob_name = audio_file_name
upload_blob(bucket_name, source_file_name, destination_blob_name)
gcs_uri = 'gs://' + bucketname + '/' + audio_file_name
transcript = ''
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code='en-IN',
enable_speaker_diarization=True,
diarization_speaker_count=2)
# Detects speech in the audio file
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
tag = 1
speaker = ""
for word_info in words_info:
if word_info.speaker_tag == tag:
speaker = speaker + " " + word_info.word
else:
transcript += f"speaker {tag}: {speaker}\n"
tag = word_info.speaker_tag
speaker = "" + word_info.word
transcript += f"speaker {tag}: {speaker}"
delete_blob(bucket_name, destination_blob_name)
return transcript
def transcribe_file(speech_file):
"""Transcribe the given audio file asynchronously."""
from google.cloud import speech_v1p1beta1
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech_v1p1beta1.SpeechClient()
# [START migration_async_request]
with io.open(speech_file, 'rb') as audio_file:
content = audio_file.read()
print "Using ", speech_file, ", with the below config:"
print ""
print "importing speech_v1p1beta1"
print "language_code='en-US'"
print "use_enhanced=True"
print "enable_automatic_punctuation=False"
print "enable_word_time_offsets=False"
print "profanity_filter=True"
print "sample_rate=44100hz"
print ""
print "Transcript is as follows"
audio = types.RecognitionAudio(content=content)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=44100,
language_code='en-US',
use_enhanced=True,
# A model must be specified to use enhanced model.
model='phone_call',
enable_automatic_punctuation=False,
enable_word_time_offsets=False,
profanity_filter=True,
# speech_contexts=[speech.types.SpeechContext(
# phrases=['Andy', 'Wisy', 'EEP', 'Project', 'Tom', 'Jeff'],
# )],
)
# [START migration_async_response]
operation = client.long_running_recognize(config, audio)
# [END migration_async_request]
print('Waiting for operation to complete...')
response = operation.result(timeout=90)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
print(u'Transcript: {}'.format(result.alternatives[0].transcript))
print('Confidence: {}'.format(result.alternatives[0].confidence))
def transcribe_streaming(stream_file, encoding="LINEAR16", sample_rate=16000):
client = speech.SpeechClient()
with io.open(stream_file, 'rb') as audio_file:
content = audio_file.read()
# In practice, stream should be a generator yielding chunks of audio data.
stream = [content]
requests = (types.StreamingRecognizeRequest(audio_content=chunk)
for chunk in stream)
config = types.RecognitionConfig(
encoding=ENCODINGS[encoding],
sample_rate_hertz=sample_rate,
language_code='ko-KR',
enable_automatic_punctuation=True,
enable_word_time_offsets=True,
enable_speaker_diarization=True, # 한국어 지원 안됨 (speaker_tag가 모두 동일인으로 분류됨)
diarization_speaker_count=3)
streaming_config = types.StreamingRecognitionConfig(config=config)
# streaming_recognize returns a generator.
responses = client.streaming_recognize(streaming_config, requests)
words_with_tags = []
transcripts = []
print("Waiting for transcribe...")
for response in responses:
for result in response.results:
alternatives = result.alternatives
for alternative in alternatives:
print(u'Transcript: {}'.format(alternative.transcript))
transcripts.append(
alternative.transcript) # punctuation 포함된 문장을 사용하기 위해 저장
for words in alternative.words:
word = words.word
start_time = round(
words.start_time.seconds +
words.start_time.nanos * 1e-9, 3)
end_time = round(
words.end_time.seconds + words.end_time.nanos * 1e-9,
3)
speaker_tag = words.speaker_tag
words_with_tags.append([
word, start_time, end_time, speaker_tag
]) # [word, start_time, end_time, speaker_tag]
print() # newline
return words_with_tags, transcripts
def process(self, loop):
"""
Audio stream recognition and result parsing
"""
#You can add speech contexts for better recognition
cap_speech_context = types.SpeechContext(**self.context)
metadata = types.RecognitionMetadata(**self.metadata)
client = speech.SpeechClient()
config = types.RecognitionConfig(encoding=self.encoding,
sample_rate_hertz=self.rate,
language_code=self.language,
speech_contexts=[
cap_speech_context,
],
enable_automatic_punctuation=True,
model=self.model,
metadata=metadata)
streaming_config = types.StreamingRecognitionConfig(
config=config,
interim_results=self.interim_results,
single_utterance=self.single_utterance)
audio_generator = self.stream_generator()
requests = iter(
types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = client.streaming_recognize(streaming_config, requests)
#print('process',type(responses))
try:
#print('process')
for response in responses:
#print('process received')
if self.terminated:
break
if not response.results:
continue
result = response.results[0]
if not result.alternatives:
continue
speechData = MessageToDict(response)
global_async_worker.add_task(self.async_callback(speechData))
# debug
transcript = result.alternatives[0].transcript
print('>>', transcript, "(OK)" if result.is_final else "")
except Exception as e:
print('process excepted', e)
self.start()
def sample_long_running_recognize(local_file_path):
"""
Print confidence level for individual words in a transcription of a short audio
file
Separating different speakers in an audio file recording
Args:
local_file_path Path to local audio file, e.g. /path/audio.wav
"""
client = speech_v1p1beta1.SpeechClient()
# local_file_path = 'audio_files/2speaker.m4a'
# If enabled, each word in the first alternative of each result will be
# tagged with a speaker tag to identify the speaker.
enable_speaker_diarization = True
# Optional. Specifies the estimated number of speakers in the conversation.
diarization_speaker_count = 2
# The language of the supplied audio
language_code = "en-US"
config = types.RecognitionConfig(
enable_speaker_diarization=enable_speaker_diarization,
enable_automatic_punctuation=True,
#diarization_speaker_count=diarization_speaker_count,
# model='phone_call',
language_code='en-US')
with io.open(local_file_path, "rb") as f:
content = f.read()
audio = types.RecognitionAudio(content=content)
operation = client.long_running_recognize(config, audio)
print(u"Waiting for operation to complete...")
response = operation.result()
print(response.results)
for result in response.results:
# First alternative has words tagged with speakers
alternative = result.alternatives[0]
print(u"Transcript: {}".format(alternative.transcript))
# Print the speaker_tag of each word
for word in alternative.words:
print(u"Word: {}".format(word.word))
print(u"Speaker tag: {}".format(word.speaker_tag))
def gspeech_client(self):
"""Creates the Google Speech API client, configures it, and sends/gets
audio/text data for parsing.
"""
language_code = 'en-US'
# Hints for the API
context = types.SpeechContext(phrases=self.context)
client = speech.SpeechClient()
# Create metadata object, helps processing
metadata = types.RecognitionMetadata()
# Interaction Type:
# VOICE_SEARCH: Transcribe spoken questions and queries into text.
# VOICE_COMMAND: Transcribe voice commands, such as for controlling a device.
metadata.interaction_type = (
enums.RecognitionMetadata.InteractionType.VOICE_COMMAND)
# Microphone Distance:
# NEARFIELD: The audio was captured from a closely placed microphone.
# MIDFIELD: The speaker is within 3 meters of the microphone.
# FARFIELD: The speaker is more than 3 meters away from the microphone.
metadata.microphone_distance = (
enums.RecognitionMetadata.MicrophoneDistance.MIDFIELD)
# Device Type:
# PC: Speech was recorded using a personal computer or tablet.
# VEHICLE: Speech was recorded in a vehicle.
# OTHER_OUTDOOR_DEVICE: Speech was recorded outdoors.
# OTHER_INDOOR_DEVICE: Speech was recorded indoors.
metadata.recording_device_type = (
enums.RecognitionMetadata.RecordingDeviceType.PC)
# Media Type:
# AUDIO: The speech data is an audio recording.
# VIDEO: The speech data originally recorded on a video.
metadata.original_media_type = (
enums.RecognitionMetadata.OriginalMediaType.AUDIO)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=16000,
language_code=language_code,
speech_contexts=[context],
use_enhanced=True,
model='command_and_search',
metadata=metadata)
streaming_config = types.StreamingRecognitionConfig(
config=config, single_utterance=False, interim_results=False)
# Hack from Google Speech Python docs, very pythonic c:
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in self.generator())
responses = client.streaming_recognize(streaming_config, requests)
self._listen_print_loop(responses)
def __init__(self):
self.client = speech.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=16000,
language_code='en-US',
enable_word_time_offsets=True,
model='video',
diarization_speaker_count=2,
enable_automatic_punctuation=True,
use_enhanced=True,
enable_speaker_diarization=True,
speech_contexts=[speech.types.SpeechContext(phrases=[])]
)
self.streaming_config = types.StreamingRecognitionConfig(config=config)
def google_transcribe(uploaded_file_path):
print("Converting: \t" + uploaded_file_path.split("/")[-1])
wav_file_path = mp3_to_wav(uploaded_file_path)
print("Converted: \t" + wav_file_path.split("/")[-1])
print("Checking frame rate: \t", wav_file_path.split("/")[-1])
frame_rate, channels = frame_rate_channel(wav_file_path)
wav_name = wav_file_path.split("/")[-1]
print("Uploading blob: \t", wav_name)
upload_blob(bucket_name, wav_file_path, wav_name)
print("Starting Transcripting: \t", wav_name)
gcs_uri = 'gs://' + bucket_name + '/' + wav_name
transcript = ''
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code=Language_code,
enable_speaker_diarization=True,
diarization_speaker_count=2)
# Detects speech in the audio file
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
tag = 1
speaker = ""
for word_info in words_info:
if word_info.speaker_tag == tag:
speaker = speaker + " " + word_info.word
else:
transcript += "speaker {}: {}".format(tag, speaker) + '\n'
tag = word_info.speaker_tag
speaker = "" + word_info.word
transcript += "speaker {}: {}".format(tag, speaker)
print("Deleting blob: \t", wav_name)
delete_blob(bucket_name, wav_name)
return transcript
def listen(self, language_code='ja-JP'):
"""Listen."""
# See http://g.co/cloud/speech/docs/languages
# for a list of supported languages.
client = speech.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=self.rate,
model=None,
speech_contexts=[types.SpeechContext(
)],
language_code=language_code)
streaming_config = types.StreamingRecognitionConfig(
config=config,
single_utterance=True,
interim_results=True
)
self.callbacks.get("ready", lambda: True)()
with MicrophoneStream(self.rate, int(self.rate/10)) as stream:
self.callbacks.get("start", lambda: True)()
while True:
try:
audio_generator = stream.generator()
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = client.streaming_recognize(streaming_config, requests)
self.listen_print_loop(responses)
except exceptions.OutOfRange:
print("Time exceeded.(OutOfRange)")
except exceptions.ServiceUnavailable:
print("Connection closed.(ServiceUnavailable)")
except KeyboardInterrupt:
print("KeyboardInterrupt.")
break
except:
print("Unexpected error:", sys.exc_info()[0])
raise
self.callbacks.get("end", lambda: True)()
def __init__(self, speakers, speaker_count, sample_rate, chunk, language_code, exit_command):
self.speakers = speakers
self.speaker_count = speaker_count
self.sample_rate = sample_rate
self.chunk = chunk
self.language_code = language_code
self.exit_command = exit_command
self.client = speech.SpeechClient()
self.recognition_config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=self.sample_rate,
language_code=self.language_code,
enable_speaker_diarization=True,
diarization_speaker_count=self.speaker_count)
self.streaming_config = types.StreamingRecognitionConfig(
config=self.recognition_config,
interim_results=True)