Alyssa Smith revised this gist . Go to revision
1 file changed, 113 insertions
transcribe_google_speech_to_text.py(file created)
| @@ -0,0 +1,113 @@ | |||
| 1 | + | #!/usr/bin/env python3 | |
| 2 | + | # pip3 install google_cloud_storage google_cloud_speech srt | |
| 3 | + | from google.cloud import speech, storage | |
| 4 | + | from sys import stderr | |
| 5 | + | from time import sleep | |
| 6 | + | from argparse import ArgumentParser | |
| 7 | + | ||
| 8 | + | parser = ArgumentParser() | |
| 9 | + | parser.add_argument("filename") | |
| 10 | + | parser.add_argument("-w", "--word_time", action="store_true", help="enable_word_time_offsets") | |
| 11 | + | parser.add_argument("-p", "--punctuation", action="store_true") | |
| 12 | + | parser.add_argument("-l", "--lang", default="en_US") | |
| 13 | + | parser.add_argument("-r", "--sample_rate", type=int, default=44100) | |
| 14 | + | parser.add_argument("--bucket", default="adslivetranscribe") | |
| 15 | + | parser.add_argument("--keep_in_storage", action="store_true") | |
| 16 | + | args = parser.parse_args() | |
| 17 | + | if args.word_time: | |
| 18 | + | import srt, datetime | |
| 19 | + | ||
| 20 | + | # from https://medium.com/searce/generate-srt-file-subtitles-using-google-clouds-speech-to-text-api-402b2f1da3bd | |
| 21 | + | def subtitle_generation(response, bin_size=3): | |
| 22 | + | """We define a bin of time period to display the words in sync with audio. | |
| 23 | + | Here, bin_size = 3 means each bin is of 3 secs. | |
| 24 | + | All the words in the interval of 3 secs in result will be grouped togather.""" | |
| 25 | + | transcriptions = [] | |
| 26 | + | index = 0 | |
| 27 | + | ||
| 28 | + | for result in response.results: | |
| 29 | + | try: | |
| 30 | + | if result.alternatives[0].words[0].start_time.seconds: | |
| 31 | + | # bin start -> for first word of result | |
| 32 | + | start_sec = result.alternatives[0].words[0].start_time.seconds | |
| 33 | + | start_microsec = result.alternatives[0].words[0].start_time.microseconds | |
| 34 | + | else: | |
| 35 | + | # bin start -> For First word of response | |
| 36 | + | start_sec = 0 | |
| 37 | + | start_microsec = 0 | |
| 38 | + | end_sec = start_sec + bin_size # bin end sec | |
| 39 | + | ||
| 40 | + | # for last word of result | |
| 41 | + | last_word_end_sec = result.alternatives[0].words[-1].end_time.seconds | |
| 42 | + | last_word_end_microsec = result.alternatives[0].words[-1].end_time.microseconds | |
| 43 | + | ||
| 44 | + | # bin transcript | |
| 45 | + | transcript = result.alternatives[0].words[0].word | |
| 46 | + | ||
| 47 | + | index += 1 # subtitle index | |
| 48 | + | ||
| 49 | + | for i in range(len(result.alternatives[0].words) - 1): | |
| 50 | + | try: | |
| 51 | + | word = result.alternatives[0].words[i + 1].word | |
| 52 | + | word_start_sec = result.alternatives[0].words[i + 1].start_time.seconds | |
| 53 | + | word_start_microsec = result.alternatives[0].words[i + 1].start_time.microseconds # 0.001 to convert nana -> micro | |
| 54 | + | word_end_sec = result.alternatives[0].words[i + 1].end_time.seconds | |
| 55 | + | word_end_microsec = result.alternatives[0].words[i + 1].end_time.microseconds | |
| 56 | + | ||
| 57 | + | if word_end_sec < end_sec: | |
| 58 | + | transcript = transcript + " " + word | |
| 59 | + | else: | |
| 60 | + | previous_word_end_sec = result.alternatives[0].words[i].end_time.seconds | |
| 61 | + | previous_word_end_microsec = result.alternatives[0].words[i].end_time.microseconds | |
| 62 | + | ||
| 63 | + | # append bin transcript | |
| 64 | + | transcriptions.append(srt.Subtitle(index, datetime.timedelta(0, start_sec, start_microsec), datetime.timedelta(0, previous_word_end_sec, previous_word_end_microsec), transcript)) | |
| 65 | + | ||
| 66 | + | # reset bin parameters | |
| 67 | + | start_sec = word_start_sec | |
| 68 | + | start_microsec = word_start_microsec | |
| 69 | + | end_sec = start_sec + bin_size | |
| 70 | + | transcript = result.alternatives[0].words[i + 1].word | |
| 71 | + | ||
| 72 | + | index += 1 | |
| 73 | + | except IndexError: | |
| 74 | + | pass | |
| 75 | + | # append transcript of last transcript in bin | |
| 76 | + | transcriptions.append(srt.Subtitle(index, datetime.timedelta(0, start_sec, start_microsec), datetime.timedelta(0, last_word_end_sec, last_word_end_microsec), transcript)) | |
| 77 | + | index += 1 | |
| 78 | + | except IndexError: | |
| 79 | + | pass | |
| 80 | + | ||
| 81 | + | # turn transcription list into subtitles | |
| 82 | + | subtitles = srt.compose(transcriptions) | |
| 83 | + | return subtitles | |
| 84 | + | ||
| 85 | + | client = speech.SpeechClient() | |
| 86 | + | ||
| 87 | + | config = speech.RecognitionConfig(encoding=speech.RecognitionConfig.AudioEncoding.ENCODING_UNSPECIFIED, sample_rate_hertz=args.sample_rate, language_code=args.lang, enable_automatic_punctuation=args.punctuation, enable_word_time_offsets=args.word_time) | |
| 88 | + | ||
| 89 | + | storage_client = storage.Client() | |
| 90 | + | bucket = storage_client.bucket(args.bucket) | |
| 91 | + | blob = bucket.blob(args.filename) | |
| 92 | + | print("uploading {}...".format(blob.name), file=stderr) | |
| 93 | + | blob.upload_from_filename(blob.name) | |
| 94 | + | print("done uploading, processing", file=stderr) | |
| 95 | + | audio = speech.RecognitionAudio(uri="gs://{}/{}".format(args.bucket, blob.name)) | |
| 96 | + | ||
| 97 | + | operation = client.long_running_recognize(config=config, audio=audio) | |
| 98 | + | ||
| 99 | + | x = 0 | |
| 100 | + | while not operation.done(): | |
| 101 | + | print("Waiting" + ("." * x) + "\r", end="", file=stderr) | |
| 102 | + | x += 1 | |
| 103 | + | sleep(2) | |
| 104 | + | ||
| 105 | + | print("", file=stderr) | |
| 106 | + | response = operation.result() | |
| 107 | + | ||
| 108 | + | if not args.word_time: | |
| 109 | + | print("".join(r.alternatives[0].transcript for r in response.results)) | |
| 110 | + | else: | |
| 111 | + | print(subtitle_generation(response)) | |
| 112 | + | if not args.keep_in_storage: | |
| 113 | + | blob.delete() | |
Newer
Older