CODEWITHSUNDEEP
pythongoogle-apispeech-to-text
tmo
tmo

Reputation: 1543

Google cloud speech to text in Python: Save translation and time to JSON

I am using the standard solution to do speech to text processing with time stamps (see code below). I know from this post that it is possible to add arguments to the gcloud commandline tool, like --format=json.

General question: How do I specify those in google.cloud.speech? I can't seem to find any documentation on Googles site on how to do this with Python.

Specific question: My aim right now, is to write out a dictionary style JSON file that contains entries for all words, plus their start and end time per word. I realise that I cloud write a hacky solution, but if a option already exists, that would be preferable.

Code:

def transcribe_file_with_word_time_offsets(speech_file, language):
    """Transcribe the given audio file synchronously and output the word time
    offsets."""
    print("Start")

    from google.cloud import speech
    from google.cloud.speech import enums
    from google.cloud.speech import types

    print("checking credentials")

    client = speech.SpeechClient(credentials=credentials)

    print("Checked")
    with io.open(speech_file, 'rb') as audio_file:
        content = audio_file.read()


    print("audio file read")

    audio = types.RecognitionAudio(content=content)

    print("config start")
    config = types.RecognitionConfig(
            encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
            language_code=language,
            enable_word_time_offsets=True)

    print("Recognizing:")
    response = client.recognize(config, audio)
    print("Recognized")

    for result in response.results:
        alternative = result.alternatives[0]
        print('Transcript: {}'.format(alternative.transcript))

        for word_info in alternative.words:
            word = word_info.word
            start_time = word_info.start_time
            end_time = word_info.end_time
            print('Word: {}, start_time: {}, end_time: {}'.format(
                word,
                start_time.seconds + start_time.nanos * 1e-9,
                end_time.seconds + end_time.nanos * 1e-9))

if __name__ == '__main__':
    parser = argparse.ArgumentParser(description=__doc__,
        formatter_class=argparse.RawDescriptionHelpFormatter)
    parser.add_argument(dest='path', help='Audio file to be recognized')
    args = parser.parse_args()
    transcribe_file_with_word_time_offsets(args.path, 'en-US')

And here is the hacky solution:

...
    transcript_dict = {'Word':[], 'start_time': [], 'end_time':[]}

    for result in response.results:
        alternative = result.alternatives[0]
        print('Transcript: {}'.format(alternative.transcript))

        for word_info in alternative.words:
            word = word_info.word
            start_time = word_info.start_time
            end_time = word_info.end_time
            transcript_dict['Word'].append(word)
            transcript_dict['start_time'].append(
                start_time.seconds + start_time.nanos * 1e-9)
            transcript_dict['end_time'].append(
                end_time.seconds + end_time.nanos * 1e-9)

    print(transcript_dict)
...

Upvotes: 6

Views: 1725

Answers (2)

Adrien Renaud
Adrien Renaud

Reputation: 2787

You could try something like:

from google.cloud import speech_v1p1beta1 as speech
import proto

client = speech.SpeechClient()

audio = speech.RecognitionAudio(...)
config = speech.RecognitionConfig(...)
operation = client.long_running_recognize(config=config, audio=audio)
response = operation.result()

response_dict = proto.Message.to_dict(response)

Upvotes: 0

Anne
Anne

Reputation: 273

The solutions using protobuf in the linked question didn't work for me (November 2020), but it led me to this comment, which worked for me with the Speech API:

speech.types.RecognizeResponse.to_json(response)

# alternatively
type(response).to_json(response)

Example

from google.cloud import speech_v1 as speech


def transcribe_gcs(gcs_uri):
    client = speech.SpeechClient()

    audio = speech.RecognitionAudio(uri=gcs_uri)
    config = speech.RecognitionConfig(
        language_code="en-US",
    )

    return client.recognize(config=config, audio=audio)


sample_audio_uri = "gs://cloud-samples-tests/speech/brooklyn.flac"

response = transcribe_gcs(sample_audio_uri)
response_json = type(response).to_json(response)


print(response_json)

{
  "results": [
    {
      "alternatives": [
        {
          "transcript": "how old is the Brooklyn Bridge",
          "confidence": 0.98314303,
          "words": []
        }
      ],
      "channelTag": 0
    }
  ]
}

Upvotes: 5

Related Questions