8156742544

Committed 05 Mar 2024 01:03PM UTC coverage: 69.889% (-0.3%) from 70.212%

Build # 8156742544

Build Type

push

github

Committed by

web-flow

Commit Message

Merge pull request #1054 from EsupPortail/develop

[DONE] Develop #3.5.1

Run Details

281 of 580 new or added lines in 23 files covered. (48.45%)

19 existing lines in 11 files now uncovered.

9911 of 14181 relevant lines covered (69.89%)

0.7 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

0.0

/pod/video_encode_transcript/transcript_model.py

import numpy as np
import shlex
import subprocess
import json

import sys
import os
from timeit import default_timer as timer
import datetime as dt
from datetime import timedelta

from webvtt import WebVTT, Caption

try:
    from shhlex import quote
except ImportError:
    from pipes import quote

import logging

try:
    from ..custom import settings_local
except ImportError:
    from .. import settings as settings_local

from .encoding_utils import sec_to_timestamp

DEBUG = getattr(settings_local, "DEBUG", False)

TRANSCRIPTION_MODEL_PARAM = getattr(settings_local, "TRANSCRIPTION_MODEL_PARAM", False)
USE_TRANSCRIPTION = getattr(settings_local, "USE_TRANSCRIPTION", False)
if USE_TRANSCRIPTION:
    TRANSCRIPTION_TYPE = getattr(settings_local, "TRANSCRIPTION_TYPE", "VOSK")
    if TRANSCRIPTION_TYPE == "VOSK":
        from vosk import Model, KaldiRecognizer
    elif TRANSCRIPTION_TYPE == "STT":
        from stt import Model
    elif TRANSCRIPTION_TYPE == "WHISPER":
        import whisper

TRANSCRIPTION_NORMALIZE = getattr(settings_local, "TRANSCRIPTION_NORMALIZE", False)
TRANSCRIPTION_NORMALIZE_TARGET_LEVEL = getattr(
    settings_local, "TRANSCRIPTION_NORMALIZE_TARGET_LEVEL", -16.0
)

TRANSCRIPTION_AUDIO_SPLIT_TIME = getattr(
    settings_local, "TRANSCRIPTION_AUDIO_SPLIT_TIME", 600
)  # 10min
# time in sec for phrase length
TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH = getattr(
    settings_local, "TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH", 2
)
TRANSCRIPTION_STT_SENTENCE_BLANK_SPLIT_TIME = getattr(
    settings_local, "TRANSCRIPTION_STT_SENTENCE_BLANK_SPLIT_TIME", 0.5
)
log = logging.getLogger(__name__)


def get_model(lang):
    """Get model for STT or Vosk software to transcript audio."""
    transript_model = Model(TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["model"])
    if TRANSCRIPTION_TYPE == "STT":
        if TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get("beam_width"):
            transript_model.setBeamWidth(
                TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["beam_width"]
            )
        if TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get("scorer"):
            print(
                "Loading scorer from files {}".format(
                    TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["scorer"]
                ),
                file=sys.stderr,
            )
            scorer_load_start = timer()
            transript_model.enableExternalScorer(
                TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["scorer"]
            )
            scorer_load_end = timer() - scorer_load_start
            print("Loaded scorer in {:.3}s.".format(scorer_load_end), file=sys.stderr)
            if TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get(
                "lm_alpha"
            ) and TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get("lm_beta"):
                transript_model.setScorerAlphaBeta(
                    TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["lm_alpha"],
                    TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["lm_beta"],
                )
    return transript_model


def start_transcripting(mp3filepath, duration, lang):
    """
    Start direct transcription.

    Normalize the audio if set, get the model according to the lang and start transcript.
    """
    if TRANSCRIPTION_NORMALIZE:
        mp3filepath = normalize_mp3(mp3filepath)
    if TRANSCRIPTION_TYPE == "WHISPER":
        msg, webvtt, all_text = main_whisper_transcript(mp3filepath, duration, lang)
    else:
        transript_model = get_model(lang)
        msg, webvtt, all_text = start_main_transcript(
            mp3filepath, duration, transript_model
        )
    if DEBUG:
        print(msg)
        print(webvtt)
        print("\n%s\n" % all_text)

    return msg, webvtt


def start_main_transcript(mp3filepath, duration, transript_model):
    """Call transcription depending software type."""
    if TRANSCRIPTION_TYPE == "STT":
        msg, webvtt, all_text = main_stt_transcript(
            mp3filepath, duration, transript_model
        )
    elif TRANSCRIPTION_TYPE == "VOSK":
        msg, webvtt, all_text = main_vosk_transcript(
            mp3filepath, duration, transript_model
        )
    return msg, webvtt, all_text


def convert_samplerate(audio_path, desired_sample_rate, trim_start, duration):
    """Convert audio to subaudio and add good sample rate."""
    sox_cmd = "sox {} --type raw --bits 16 --channels 1 --rate {} ".format(
        quote(audio_path), desired_sample_rate
    )
    sox_cmd += "--encoding signed-integer --endian little --compression 0.0 "
    sox_cmd += "--no-dither - trim {} {}".format(trim_start, duration)

    try:
        output = subprocess.check_output(shlex.split(sox_cmd), stderr=subprocess.PIPE)

    except subprocess.CalledProcessError as e:
        raise RuntimeError("SoX returned non-zero status: {}".format(e.stderr))
    except OSError as e:
        raise OSError(
            e.errno,
            "SoX not found, use {}hz files or install it: {}".format(
                desired_sample_rate, e.strerror
            ),
        )

    if TRANSCRIPTION_TYPE == "WHISPER":
        return np.frombuffer(output, np.int16).flatten().astype(np.float32) / 32768.0
    else:
        return np.frombuffer(output, np.int16)


def normalize_mp3(mp3filepath):
    """Normalize the audio to good format and sound level."""
    filename, file_extension = os.path.splitext(mp3filepath)
    mp3normfile = "{}{}{}".format(filename, "_norm", file_extension)
    normalize_cmd = "ffmpeg-normalize {} ".format(quote(mp3filepath))
    normalize_cmd += "-c:a libmp3lame -b:a 192k --normalization-type ebu "
    # normalize_cmd += \
    # '--loudness-range-target 7.0 --true-peak 0.0 --offset 0.0 '
    normalize_cmd += "--target-level {} -f -o {}".format(
        TRANSCRIPTION_NORMALIZE_TARGET_LEVEL, quote(mp3normfile)
    )
    if DEBUG:
        print(normalize_cmd)
    try:
        subprocess.check_output(shlex.split(normalize_cmd), stderr=subprocess.PIPE)
        return mp3normfile
    except subprocess.CalledProcessError as e:
        log.error("ffmpeg-normalize returned non-zero status: {}".format(e.stderr))
        return mp3filepath
    except OSError as e:
        log.error("ffmpeg-normalize not found {}".format(e.strerror))
        return mp3filepath


# #################################
# TRANSCRIPT VIDEO: MAIN FUNCTION
# #################################


def convert_vosk_samplerate(audio_path, desired_sample_rate, trim_start, duration):
    """Convert audio to the good sample rate."""
    sox_cmd = "sox {} --type raw --bits 16 --channels 1 --rate {} ".format(
        quote(audio_path), desired_sample_rate
    )
    sox_cmd += "--encoding signed-integer --endian little --compression 0.0 "
    sox_cmd += "--no-dither - trim {} {}".format(trim_start, duration)

    try:
        output = subprocess.Popen(shlex.split(sox_cmd), stdout=subprocess.PIPE)

    except subprocess.CalledProcessError as e:
        raise RuntimeError("SoX returned non-zero status: {}".format(e.stderr))
    except OSError as e:
        raise OSError(
            e.errno,
            "SoX not found, use {}hz files or install it: {}".format(
                desired_sample_rate, e.strerror
            ),
        )
    return output


def get_word_result_from_data(results, audio, rec):
    """Get subsound from audio and add transcription to result parameter."""
    while True:
        data = audio.stdout.read(4000)
        if len(data) == 0:
            break
        if rec.AcceptWaveform(data):
            results.append(rec.Result())
    results.append(rec.Result())


def words_to_vtt(
    words,
    start_trim,
    duration,
    is_first_caption,
    text_caption,
    start_caption,
    last_word_added,
    all_text,
    webvtt,
):
    """Convert word and time to webvtt captions."""
    for index, word in enumerate(words):
        start_key = "start_time"
        word_duration = word.get("duration", 0)
        last_word = words[-1]
        last_word_duration = last_word.get("duration", 0)
        if TRANSCRIPTION_TYPE == "VOSK":
            start_key = "start"
            word_duration = word["end"] - word["start"]
            last_word_duration = words[-1]["end"] - words[-1]["start"]
        next_word = None
        blank_duration = 0
        if word != words[-1] and (index + 1) < len(words):
            next_word = words[index + 1]
            blank_duration = ((next_word[start_key]) - start_caption) - (
                ((word[start_key]) - start_caption) + word_duration
            )
        all_text += word["word"] + " "
        # word: <class 'dict'> {'word': 'bonjour', 'start ':
        # 0.58, 'duration': 7.34}
        text_caption.append(word["word"])
        if not (
            (((word[start_key]) - start_caption) < TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH)
            and (
                next_word is not None
                and (blank_duration < TRANSCRIPTION_STT_SENTENCE_BLANK_SPLIT_TIME)
            )
        ):
            # on créé le caption
            if is_first_caption:
                # A revoir, fusion de la nouvelle ligne avec
                # l'ancienne...
                is_first_caption = False
                text_caption = get_text_caption(text_caption, last_word_added)

            stop_caption = word[start_key] + word_duration

            # on evite le chevauchement
            change_previous_end_caption(webvtt, start_caption)

            caption = Caption(
                sec_to_timestamp(start_caption),
                sec_to_timestamp(stop_caption),
                " ".join(text_caption),
            )

            webvtt.captions.append(caption)
            # on remet tout à zero pour la prochaine phrase
            start_caption = word[start_key]
            text_caption = []
            last_word_added = word["word"]
    if start_trim + TRANSCRIPTION_AUDIO_SPLIT_TIME > duration:
        # on ajoute ici la dernière phrase de la vidéo
        stop_caption = start_trim + words[-1][start_key] + last_word_duration
        caption = Caption(
            sec_to_timestamp(start_caption),
            sec_to_timestamp(stop_caption),
            " ".join(text_caption),
        )
        webvtt.captions.append(caption)
    return all_text, webvtt


def main_vosk_transcript(norm_mp3_file, duration, transript_model):
    """Vosk transcription."""
    msg = ""
    inference_start = timer()
    msg += "\nInference start %0.3fs." % inference_start
    desired_sample_rate = 16000

    rec = KaldiRecognizer(transript_model, desired_sample_rate)
    rec.SetWords(True)

    webvtt = WebVTT()
    all_text = ""
    for start_trim in range(0, duration, TRANSCRIPTION_AUDIO_SPLIT_TIME):
        audio = convert_vosk_samplerate(
            norm_mp3_file,
            desired_sample_rate,
            start_trim,
            TRANSCRIPTION_AUDIO_SPLIT_TIME,  # dur
        )
        msg += "\nRunning inference."
        results = []
        get_word_result_from_data(results, audio, rec)
        for res in results:
            words = json.loads(res).get("result")
            text = json.loads(res).get("text")
            if not words:
                continue
            start_caption = words[0]["start"]
            stop_caption = words[-1]["end"]
            caption = Caption(
                sec_to_timestamp(start_caption),
                sec_to_timestamp(stop_caption),
                text,
            )
            webvtt.captions.append(caption)
            """
            text_caption = []
            is_first_caption = True
            all_text, webvtt = words_to_vtt(
                words,
                start_trim,
                duration,
                is_first_caption,
                text_caption,
                start_caption,
                last_word_added,
                all_text,
                webvtt,
            )
            """
    inference_end = timer() - inference_start

    msg += "\nInference took %0.3fs." % inference_end
    return msg, webvtt, all_text


def main_stt_transcript(norm_mp3_file, duration, transript_model):
    """STT transcription."""
    msg = ""
    inference_start = timer()
    msg += "\nInference start %0.3fs." % inference_start
    desired_sample_rate = transript_model.sampleRate()
    webvtt = WebVTT()
    last_word_added = ""
    metadata = None
    all_text = ""
    for start_trim in range(0, duration, TRANSCRIPTION_AUDIO_SPLIT_TIME):
        end_trim = (
            duration
            if start_trim + TRANSCRIPTION_AUDIO_SPLIT_TIME > duration
            else (
                start_trim
                + TRANSCRIPTION_AUDIO_SPLIT_TIME
                + TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH
            )
        )

        dur = (
            (TRANSCRIPTION_AUDIO_SPLIT_TIME + TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH)
            if (
                (
                    start_trim
                    + TRANSCRIPTION_AUDIO_SPLIT_TIME
                    + TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH
                )
                < duration
            )
            else (duration - start_trim)
        )

        msg += "\ntake audio from %s to %s - %s" % (start_trim, end_trim, dur)

        audio = convert_samplerate(norm_mp3_file, desired_sample_rate, start_trim, dur)
        msg += "\nRunning inference."

        metadata = transript_model.sttWithMetadata(audio)

        for transcript in metadata.transcripts:
            msg += "\nConfidence: %s" % transcript.confidence
            words = words_from_candidate_transcript(transcript)
            start_caption = start_trim + words[0]["start_time"]
            text_caption = []
            is_first_caption = True
            all_text, webvtt = words_to_vtt(
                words,
                start_trim,
                duration,
                is_first_caption,
                text_caption,
                start_caption,
                last_word_added,
                all_text,
                webvtt,
            )
    inference_end = timer() - inference_start

    msg += "\nInference took %0.3fs." % inference_end
    return msg, webvtt, all_text


def main_whisper_transcript(norm_mp3_file, duration, lang):
    """Whisper transcription."""
    msg = ""
    all_text = ""
    webvtt = WebVTT()
    inference_start = timer()
    desired_sample_rate = 16000
    msg += "\nInference start %0.3fs." % inference_start

    model = whisper.load_model(
        TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["model"],
        download_root=TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang][
            "download_root"
        ],
    )

    for start_trim in range(0, duration, TRANSCRIPTION_AUDIO_SPLIT_TIME):
        log.info("start_trim: " + str(start_trim))
        audio = convert_samplerate(
            norm_mp3_file,
            desired_sample_rate,
            start_trim,
            TRANSCRIPTION_AUDIO_SPLIT_TIME,  # dur
        )
        transcription = model.transcribe(audio, language=lang)
        msg += "\nRunning inference."
        for segment in transcription["segments"]:
            caption = Caption(
                sec_to_timestamp(segment["start"] + start_trim),
                sec_to_timestamp(segment["end"] + start_trim),
                segment["text"],
            )
            webvtt.captions.append(caption)

    inference_end = timer() - inference_start
    msg += "\nInference took %0.3fs." % inference_end
    return msg, webvtt, all_text


def change_previous_end_caption(webvtt, start_caption):
    """Change the end time for caption."""
    if len(webvtt.captions) > 0:
        prev_end = dt.datetime.strptime(webvtt.captions[-1].end, "%H:%M:%S.%f")
        td_prev_end = timedelta(
            hours=prev_end.hour,
            minutes=prev_end.minute,
            seconds=prev_end.second,
            microseconds=prev_end.microsecond,
        ).total_seconds()
        if td_prev_end > start_caption:
            webvtt.captions[-1].end = sec_to_timestamp(start_caption)


def get_text_caption(text_caption, last_word_added):
    """Get the text for a caption."""
    try:
        first_index = text_caption.index(last_word_added)
        return text_caption[first_index + 1 :]
    except ValueError:
        return text_caption


def words_from_candidate_transcript(metadata):
    """Get words list from transcription."""
    word = ""
    word_list = []
    word_start_time = 0
    # Loop through each character
    for i, token in enumerate(metadata.tokens):
        # Append character to word if it's not a space
        if token.text != " ":
            if len(word) == 0:
                # Log the start time of the new word
                word_start_time = token.start_time

            word = word + token.text
        # Word boundary is either a space or the last character in the array
        if token.text == " " or i == len(metadata.tokens) - 1:
            word_duration = token.start_time - word_start_time

            if word_duration < 0:
                word_duration = 0

            each_word = dict()
            each_word["word"] = word
            each_word["start_time"] = round(word_start_time, 4)
            each_word["duration"] = round(word_duration, 4)

            word_list.append(each_word)
            # Reset
            word = ""
            word_start_time = 0

    return word_list

1	import numpy as np	×
2	import shlex	×
3	import subprocess	×
4	import json	×
5
6	import sys	×
7	import os	×
8	from timeit import default_timer as timer	×
9	import datetime as dt	×
10	from datetime import timedelta	×
11
12	from webvtt import WebVTT, Caption	×
13
14	try:	×
15	from shhlex import quote	×
16	except ImportError:	×
17	from pipes import quote	×
18
19	import logging	×
20
21	try:	×
22	from ..custom import settings_local	×
23	except ImportError:	×
24	from .. import settings as settings_local	×
25
NEW 26	from .encoding_utils import sec_to_timestamp	×
27
UNCOV 28	DEBUG = getattr(settings_local, "DEBUG", False)	×
29
30	TRANSCRIPTION_MODEL_PARAM = getattr(settings_local, "TRANSCRIPTION_MODEL_PARAM", False)	×
31	USE_TRANSCRIPTION = getattr(settings_local, "USE_TRANSCRIPTION", False)	×
32	if USE_TRANSCRIPTION:	×
33	TRANSCRIPTION_TYPE = getattr(settings_local, "TRANSCRIPTION_TYPE", "VOSK")	×
34	if TRANSCRIPTION_TYPE == "VOSK":	×
35	from vosk import Model, KaldiRecognizer	×
36	elif TRANSCRIPTION_TYPE == "STT":	×
37	from stt import Model	×
38	elif TRANSCRIPTION_TYPE == "WHISPER":	×
39	import whisper	×
40
41	TRANSCRIPTION_NORMALIZE = getattr(settings_local, "TRANSCRIPTION_NORMALIZE", False)	×
42	TRANSCRIPTION_NORMALIZE_TARGET_LEVEL = getattr(	×
43	settings_local, "TRANSCRIPTION_NORMALIZE_TARGET_LEVEL", -16.0
44	)
45
46	TRANSCRIPTION_AUDIO_SPLIT_TIME = getattr(	×
47	settings_local, "TRANSCRIPTION_AUDIO_SPLIT_TIME", 600
48	) # 10min
49	# time in sec for phrase length
50	TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH = getattr(	×
51	settings_local, "TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH", 2
52	)
53	TRANSCRIPTION_STT_SENTENCE_BLANK_SPLIT_TIME = getattr(	×
54	settings_local, "TRANSCRIPTION_STT_SENTENCE_BLANK_SPLIT_TIME", 0.5
55	)
56	log = logging.getLogger(__name__)	×
57
58
59	def get_model(lang):	×
60	"""Get model for STT or Vosk software to transcript audio."""
61	transript_model = Model(TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["model"])	×
62	if TRANSCRIPTION_TYPE == "STT":	×
63	if TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get("beam_width"):	×
64	transript_model.setBeamWidth(	×
65	TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["beam_width"]
66	)
67	if TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get("scorer"):	×
68	print(	×
69	"Loading scorer from files {}".format(
70	TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["scorer"]
71	),
72	file=sys.stderr,
73	)
74	scorer_load_start = timer()	×
75	transript_model.enableExternalScorer(	×
76	TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["scorer"]
77	)
78	scorer_load_end = timer() - scorer_load_start	×
79	print("Loaded scorer in {:.3}s.".format(scorer_load_end), file=sys.stderr)	×
80	if TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get(	×
81	"lm_alpha"
82	) and TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang].get("lm_beta"):
83	transript_model.setScorerAlphaBeta(	×
84	TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["lm_alpha"],
85	TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["lm_beta"],
86	)
87	return transript_model	×
88
89
90	def start_transcripting(mp3filepath, duration, lang):	×
91	"""
92	Start direct transcription.
93
94	Normalize the audio if set, get the model according to the lang and start transcript.
95	"""
96	if TRANSCRIPTION_NORMALIZE:	×
97	mp3filepath = normalize_mp3(mp3filepath)	×
98	if TRANSCRIPTION_TYPE == "WHISPER":	×
99	msg, webvtt, all_text = main_whisper_transcript(mp3filepath, duration, lang)	×
100	else:
101	transript_model = get_model(lang)	×
102	msg, webvtt, all_text = start_main_transcript(	×
103	mp3filepath, duration, transript_model
104	)
105	if DEBUG:	×
106	print(msg)	×
107	print(webvtt)	×
108	print("\n%s\n" % all_text)	×
109
110	return msg, webvtt	×
111
112
113	def start_main_transcript(mp3filepath, duration, transript_model):	×
114	"""Call transcription depending software type."""
115	if TRANSCRIPTION_TYPE == "STT":	×
116	msg, webvtt, all_text = main_stt_transcript(	×
117	mp3filepath, duration, transript_model
118	)
119	elif TRANSCRIPTION_TYPE == "VOSK":	×
120	msg, webvtt, all_text = main_vosk_transcript(	×
121	mp3filepath, duration, transript_model
122	)
123	return msg, webvtt, all_text	×
124
125
126	def convert_samplerate(audio_path, desired_sample_rate, trim_start, duration):	×
127	"""Convert audio to subaudio and add good sample rate."""
128	sox_cmd = "sox {} --type raw --bits 16 --channels 1 --rate {} ".format(	×
129	quote(audio_path), desired_sample_rate
130	)
131	sox_cmd += "--encoding signed-integer --endian little --compression 0.0 "	×
132	sox_cmd += "--no-dither - trim {} {}".format(trim_start, duration)	×
133
134	try:	×
135	output = subprocess.check_output(shlex.split(sox_cmd), stderr=subprocess.PIPE)	×
136
137	except subprocess.CalledProcessError as e:	×
138	raise RuntimeError("SoX returned non-zero status: {}".format(e.stderr))	×
139	except OSError as e:	×
140	raise OSError(	×
141	e.errno,
142	"SoX not found, use {}hz files or install it: {}".format(
143	desired_sample_rate, e.strerror
144	),
145	)
146
147	if TRANSCRIPTION_TYPE == "WHISPER":	×
148	return np.frombuffer(output, np.int16).flatten().astype(np.float32) / 32768.0	×
149	else:
150	return np.frombuffer(output, np.int16)	×
151
152
153	def normalize_mp3(mp3filepath):	×
154	"""Normalize the audio to good format and sound level."""
155	filename, file_extension = os.path.splitext(mp3filepath)	×
156	mp3normfile = "{}{}{}".format(filename, "_norm", file_extension)	×
157	normalize_cmd = "ffmpeg-normalize {} ".format(quote(mp3filepath))	×
158	normalize_cmd += "-c:a libmp3lame -b:a 192k --normalization-type ebu "	×
159	# normalize_cmd += \
160	# '--loudness-range-target 7.0 --true-peak 0.0 --offset 0.0 '
161	normalize_cmd += "--target-level {} -f -o {}".format(	×
162	TRANSCRIPTION_NORMALIZE_TARGET_LEVEL, quote(mp3normfile)
163	)
164	if DEBUG:	×
165	print(normalize_cmd)	×
166	try:	×
167	subprocess.check_output(shlex.split(normalize_cmd), stderr=subprocess.PIPE)	×
168	return mp3normfile	×
169	except subprocess.CalledProcessError as e:	×
170	log.error("ffmpeg-normalize returned non-zero status: {}".format(e.stderr))	×
171	return mp3filepath	×
172	except OSError as e:	×
173	log.error("ffmpeg-normalize not found {}".format(e.strerror))	×
174	return mp3filepath	×
175
176
177	# #################################
178	# TRANSCRIPT VIDEO: MAIN FUNCTION
179	# #################################
180
181
182	def convert_vosk_samplerate(audio_path, desired_sample_rate, trim_start, duration):	×
183	"""Convert audio to the good sample rate."""
184	sox_cmd = "sox {} --type raw --bits 16 --channels 1 --rate {} ".format(	×
185	quote(audio_path), desired_sample_rate
186	)
187	sox_cmd += "--encoding signed-integer --endian little --compression 0.0 "	×
188	sox_cmd += "--no-dither - trim {} {}".format(trim_start, duration)	×
189
190	try:	×
191	output = subprocess.Popen(shlex.split(sox_cmd), stdout=subprocess.PIPE)	×
192
193	except subprocess.CalledProcessError as e:	×
194	raise RuntimeError("SoX returned non-zero status: {}".format(e.stderr))	×
195	except OSError as e:	×
196	raise OSError(	×
197	e.errno,
198	"SoX not found, use {}hz files or install it: {}".format(
199	desired_sample_rate, e.strerror
200	),
201	)
202	return output	×
203
204
205	def get_word_result_from_data(results, audio, rec):	×
206	"""Get subsound from audio and add transcription to result parameter."""
207	while True:
208	data = audio.stdout.read(4000)	×
209	if len(data) == 0:	×
210	break	×
211	if rec.AcceptWaveform(data):	×
212	results.append(rec.Result())	×
213	results.append(rec.Result())	×
214
215
216	def words_to_vtt(	×
217	words,
218	start_trim,
219	duration,
220	is_first_caption,
221	text_caption,
222	start_caption,
223	last_word_added,
224	all_text,
225	webvtt,
226	):
227	"""Convert word and time to webvtt captions."""
228	for index, word in enumerate(words):	×
229	start_key = "start_time"	×
230	word_duration = word.get("duration", 0)	×
231	last_word = words[-1]	×
232	last_word_duration = last_word.get("duration", 0)	×
233	if TRANSCRIPTION_TYPE == "VOSK":	×
234	start_key = "start"	×
235	word_duration = word["end"] - word["start"]	×
236	last_word_duration = words[-1]["end"] - words[-1]["start"]	×
237	next_word = None	×
238	blank_duration = 0	×
239	if word != words[-1] and (index + 1) < len(words):	×
240	next_word = words[index + 1]	×
241	blank_duration = ((next_word[start_key]) - start_caption) - (	×
242	((word[start_key]) - start_caption) + word_duration
243	)
244	all_text += word["word"] + " "	×
245	# word: <class 'dict'> {'word': 'bonjour', 'start ':
246	# 0.58, 'duration': 7.34}
247	text_caption.append(word["word"])	×
248	if not (	×
249	(((word[start_key]) - start_caption) < TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH)
250	and (
251	next_word is not None
252	and (blank_duration < TRANSCRIPTION_STT_SENTENCE_BLANK_SPLIT_TIME)
253	)
254	):
255	# on créé le caption
256	if is_first_caption:	×
257	# A revoir, fusion de la nouvelle ligne avec
258	# l'ancienne...
259	is_first_caption = False	×
260	text_caption = get_text_caption(text_caption, last_word_added)	×
261
262	stop_caption = word[start_key] + word_duration	×
263
264	# on evite le chevauchement
265	change_previous_end_caption(webvtt, start_caption)	×
266
267	caption = Caption(	×
268	sec_to_timestamp(start_caption),
269	sec_to_timestamp(stop_caption),
270	" ".join(text_caption),
271	)
272
273	webvtt.captions.append(caption)	×
274	# on remet tout à zero pour la prochaine phrase
275	start_caption = word[start_key]	×
276	text_caption = []	×
277	last_word_added = word["word"]	×
278	if start_trim + TRANSCRIPTION_AUDIO_SPLIT_TIME > duration:	×
279	# on ajoute ici la dernière phrase de la vidéo
280	stop_caption = start_trim + words[-1][start_key] + last_word_duration	×
281	caption = Caption(	×
282	sec_to_timestamp(start_caption),
283	sec_to_timestamp(stop_caption),
284	" ".join(text_caption),
285	)
286	webvtt.captions.append(caption)	×
287	return all_text, webvtt	×
288
289
290	def main_vosk_transcript(norm_mp3_file, duration, transript_model):	×
291	"""Vosk transcription."""
292	msg = ""	×
293	inference_start = timer()	×
294	msg += "\nInference start %0.3fs." % inference_start	×
295	desired_sample_rate = 16000	×
296
297	rec = KaldiRecognizer(transript_model, desired_sample_rate)	×
298	rec.SetWords(True)	×
299
300	webvtt = WebVTT()	×
301	all_text = ""	×
302	for start_trim in range(0, duration, TRANSCRIPTION_AUDIO_SPLIT_TIME):	×
303	audio = convert_vosk_samplerate(	×
304	norm_mp3_file,
305	desired_sample_rate,
306	start_trim,
307	TRANSCRIPTION_AUDIO_SPLIT_TIME, # dur
308	)
309	msg += "\nRunning inference."	×
310	results = []	×
311	get_word_result_from_data(results, audio, rec)	×
312	for res in results:	×
313	words = json.loads(res).get("result")	×
314	text = json.loads(res).get("text")	×
315	if not words:	×
316	continue	×
317	start_caption = words[0]["start"]	×
318	stop_caption = words[-1]["end"]	×
319	caption = Caption(	×
320	sec_to_timestamp(start_caption),
321	sec_to_timestamp(stop_caption),
322	text,
323	)
324	webvtt.captions.append(caption)	×
325	"""	×
326	text_caption = []
327	is_first_caption = True
328	all_text, webvtt = words_to_vtt(
329	words,
330	start_trim,
331	duration,
332	is_first_caption,
333	text_caption,
334	start_caption,
335	last_word_added,
336	all_text,
337	webvtt,
338	)
339	"""
340	inference_end = timer() - inference_start	×
341
342	msg += "\nInference took %0.3fs." % inference_end	×
343	return msg, webvtt, all_text	×
344
345
346	def main_stt_transcript(norm_mp3_file, duration, transript_model):	×
347	"""STT transcription."""
348	msg = ""	×
349	inference_start = timer()	×
350	msg += "\nInference start %0.3fs." % inference_start	×
351	desired_sample_rate = transript_model.sampleRate()	×
352	webvtt = WebVTT()	×
353	last_word_added = ""	×
354	metadata = None	×
355	all_text = ""	×
356	for start_trim in range(0, duration, TRANSCRIPTION_AUDIO_SPLIT_TIME):	×
357	end_trim = (	×
358	duration
359	if start_trim + TRANSCRIPTION_AUDIO_SPLIT_TIME > duration
360	else (
361	start_trim
362	+ TRANSCRIPTION_AUDIO_SPLIT_TIME
363	+ TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH
364	)
365	)
366
367	dur = (	×
368	(TRANSCRIPTION_AUDIO_SPLIT_TIME + TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH)
369	if (
370	(
371	start_trim
372	+ TRANSCRIPTION_AUDIO_SPLIT_TIME
373	+ TRANSCRIPTION_STT_SENTENCE_MAX_LENGTH
374	)
375	< duration
376	)
377	else (duration - start_trim)
378	)
379
380	msg += "\ntake audio from %s to %s - %s" % (start_trim, end_trim, dur)	×
381
382	audio = convert_samplerate(norm_mp3_file, desired_sample_rate, start_trim, dur)	×
383	msg += "\nRunning inference."	×
384
385	metadata = transript_model.sttWithMetadata(audio)	×
386
387	for transcript in metadata.transcripts:	×
388	msg += "\nConfidence: %s" % transcript.confidence	×
389	words = words_from_candidate_transcript(transcript)	×
390	start_caption = start_trim + words[0]["start_time"]	×
391	text_caption = []	×
392	is_first_caption = True	×
393	all_text, webvtt = words_to_vtt(	×
394	words,
395	start_trim,
396	duration,
397	is_first_caption,
398	text_caption,
399	start_caption,
400	last_word_added,
401	all_text,
402	webvtt,
403	)
404	inference_end = timer() - inference_start	×
405
406	msg += "\nInference took %0.3fs." % inference_end	×
407	return msg, webvtt, all_text	×
408
409
410	def main_whisper_transcript(norm_mp3_file, duration, lang):	×
411	"""Whisper transcription."""
412	msg = ""	×
413	all_text = ""	×
414	webvtt = WebVTT()	×
415	inference_start = timer()	×
416	desired_sample_rate = 16000	×
417	msg += "\nInference start %0.3fs." % inference_start	×
418
419	model = whisper.load_model(	×
420	TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang]["model"],
421	download_root=TRANSCRIPTION_MODEL_PARAM[TRANSCRIPTION_TYPE][lang][
422	"download_root"
423	],
424	)
425
426	for start_trim in range(0, duration, TRANSCRIPTION_AUDIO_SPLIT_TIME):	×
427	log.info("start_trim: " + str(start_trim))	×
428	audio = convert_samplerate(	×
429	norm_mp3_file,
430	desired_sample_rate,
431	start_trim,
432	TRANSCRIPTION_AUDIO_SPLIT_TIME, # dur
433	)
434	transcription = model.transcribe(audio, language=lang)	×
435	msg += "\nRunning inference."	×
436	for segment in transcription["segments"]:	×
437	caption = Caption(	×
438	sec_to_timestamp(segment["start"] + start_trim),
439	sec_to_timestamp(segment["end"] + start_trim),
440	segment["text"],
441	)
442	webvtt.captions.append(caption)	×
443
444	inference_end = timer() - inference_start	×
445	msg += "\nInference took %0.3fs." % inference_end	×
446	return msg, webvtt, all_text	×
447
448
449	def change_previous_end_caption(webvtt, start_caption):	×
450	"""Change the end time for caption."""
451	if len(webvtt.captions) > 0:	×
452	prev_end = dt.datetime.strptime(webvtt.captions[-1].end, "%H:%M:%S.%f")	×
453	td_prev_end = timedelta(	×
454	hours=prev_end.hour,
455	minutes=prev_end.minute,
456	seconds=prev_end.second,
457	microseconds=prev_end.microsecond,
458	).total_seconds()
459	if td_prev_end > start_caption:	×
460	webvtt.captions[-1].end = sec_to_timestamp(start_caption)	×
461
462
463	def get_text_caption(text_caption, last_word_added):	×
464	"""Get the text for a caption."""
465	try:	×
466	first_index = text_caption.index(last_word_added)	×
467	return text_caption[first_index + 1 :]	×
468	except ValueError:	×
469	return text_caption	×
470
471
472	def words_from_candidate_transcript(metadata):	×
473	"""Get words list from transcription."""
474	word = ""	×
475	word_list = []	×
476	word_start_time = 0	×
477	# Loop through each character
478	for i, token in enumerate(metadata.tokens):	×
479	# Append character to word if it's not a space
480	if token.text != " ":	×
481	if len(word) == 0:	×
482	# Log the start time of the new word
483	word_start_time = token.start_time	×
484
485	word = word + token.text	×
486	# Word boundary is either a space or the last character in the array
487	if token.text == " " or i == len(metadata.tokens) - 1:	×
488	word_duration = token.start_time - word_start_time	×
489
490	if word_duration < 0:	×
491	word_duration = 0	×
492
493	each_word = dict()	×
494	each_word["word"] = word	×
495	each_word["start_time"] = round(word_start_time, 4)	×
496	each_word["duration"] = round(word_duration, 4)	×
497
498	word_list.append(each_word)	×
499	# Reset
500	word = ""	×
501	word_start_time = 0	×
502
503	return word_list	×

EsupPortail / Esup-Pod / 8156742544

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous