def get_transcripts_json(gcstorage_path,
lang,
phrase_hints=[],
speaker_count=1,
enhanced_model=None):
# transcribes audio files
def _jsonify(res):
# helper func for simplifying gcp speech client response
json = []
for section in res.results:
data = {
'transcript': section.alternatives[0].transcript,
'words': []
}
for word in section.alternative[0].words:
data['words'].append({
'word':
word.word,
'start_time':
word.start_time.total_seconds(),
'end_time':
word.end_time.total_seconds(),
'speaker_tag':
word.speaker_tag
})
json.append(data)
return json
client = speech.SpeechClient()
audio = speech.RecognitionAudio(uri=gcstorage_path)
diarize = speaker_count if speaker_count > 1 else False
print(f"Diarizing: {diarize}")
diarizationConfig = speech.SpeakerDiarizationConfig(
enable_speaker_diarization=speaker_count
if speaker_count > 1 else False, )
# if eng only, can use the optimized video model
if lang == 'en':
enhanced_model = 'video'
config = speech.RecognitionConfig(
lang_code='en-US' if lang == 'en' else lang,
enable_automatic_punctuation=True,
enable_word_time_offsets=True,
speech_contexts=[{
'phrases': phrase_hints,
'boost': 15
}],
diarization_config=diarizationConfig,
profanity_filter=True,
use_enhanced=True if enhanced_model else False,
model='video' if enhanced_model else None)
res = client.long_running_recognize(config=config, audio=audio).result()
return _jsonify(res)
def get_transcripts_json(gcsPath,
langCode,
phraseHints=[],
speakerCount=1,
enhancedModel=None):
"""Transcribes audio files.
Args:
gcsPath (String): path to file in cloud storage (i.e. "gs://audio/clip.mp4")
langCode (String): language code (i.e. "en-US", see https://cloud.google.com/speech-to-text/docs/languages)
phraseHints (String[]): list of words that are unusual but likely to appear in the audio file.
speakerCount (int, optional): Number of speakers in the audio. Only works on English. Defaults to None.
enhancedModel (String, optional): Option to use an enhanced speech model, i.e. "video"
Returns:
list | Operation.error
"""
# Helper function for simplifying Google speech client response
def _jsonify(result):
json = []
for section in result.results:
data = {
"transcript": section.alternatives[0].transcript,
"words": []
}
for word in section.alternatives[0].words:
data["words"].append({
"word":
word.word,
"start_time":
word.start_time.total_seconds(),
"end_time":
word.end_time.total_seconds(),
"speaker_tag":
word.speaker_tag
})
json.append(data)
return json
client = speech.SpeechClient()
audio = speech.RecognitionAudio(uri=gcsPath)
diarize = speakerCount if speakerCount > 1 else False
print(f"Diarizing: {diarize}")
diarizationConfig = speech.SpeakerDiarizationConfig(
enable_speaker_diarization=speakerCount
if speakerCount > 1 else False, )
# In English only, we can use the optimized video model
if langCode == "en":
enhancedModel = "video"
config = speech.RecognitionConfig(
language_code="en-US" if langCode == "en" else langCode,
enable_automatic_punctuation=True,
enable_word_time_offsets=True,
speech_contexts=[{
"phrases": phraseHints,
"boost": 15
}],
diarization_config=diarizationConfig,
profanity_filter=True,
use_enhanced=True if enhancedModel else False,
model="video" if enhancedModel else None)
res = client.long_running_recognize(config=config, audio=audio).result()
return _jsonify(res)
from google.cloud import speech_v1p1beta1 as speech
client = speech.SpeechClient()
speech_file = "'/home/serkhane/Repositories/AI/DATA/youtube_data_taflowtron/en/jocko_podcast_shortlist/v4_concate_removesilence/1HhXDprzf5I/clips/1HhXDprzf5I_trim_0.0_6180.0.wav'"
with open(speech_file, "rb") as audio_file:
content = audio_file.read()
audio = speech.RecognitionAudio(content=content)
diarization_config = speech.SpeakerDiarizationConfig(
enable_speaker_diarization=True,
min_speaker_count=2,
max_speaker_count=10,
)
config = speech.RecognitionConfig(
encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=8000,
language_code="en-US",
diarization_config=diarization_config,
)
print("Waiting for operation to complete...")
response = client.recognize(config=config, audio=audio)
# The transcript within each result is separate and sequential per result.
# However, the words list within an alternative includes all the words
# from all the results thus far. Thus, to get all the words with speaker
# tags, you only have to take the words list from the last result: