def convert_audio_file_if_necessary(aft):
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(aft)
# Check that audio is in the right format
command = [
'ffprobe', '-v', 'quiet', '-print_format', 'json', '-show_format',
'-show_streams', tmp_file
]
p = Popen(command, stdin=PIPE, stdout=PIPE)
output, errors = p.communicate()
results = ast.literal_eval(output)
convert = False
for stream in results['streams']:
if stream["codec_type"] in 'audio':
if stream["codec_name"] not in "aac mp3 wav":
convert = True
if stream["channels"] > 2:
convert = True
if convert:
new_file = '.'.join(tmp_file.split('.')[:-1]) + '.m4a'
logger.debug(new_file)
command = [
'ffmpeg', '-i', tmp_file, '-ar', '44100', '-ac', '1', '-c:a',
'aac', new_file
]
logger.debug(command)
p = Popen(command, stdin=PIPE, stdout=PIPE)
output, errors = p.communicate()
if not errors:
aft.audio_file.save(
new_file.split('/').pop(), File(open(new_file)))
aft.save()
def create_and_return_transcription_segments(aft):
'''
Creates the transcription segments from an AudioFileTranscription model.
'''
convert_audio_file_if_necessary(aft)
# We should delete all segments if we're going to create more!
deleted = TranscriptionSegment.objects.filter(parent=aft).delete()
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(aft)
segments = wahi_korero_segmenter(tmp_file, aft)
# segments = dummy_segme nter(tmp_file)
logger.debug(segments)
ts_segments = []
for segment in segments:
start = segment['start']
end = segment['end']
ts, created = TranscriptionSegment.objects.get_or_create(start=start,
end=end,
parent=aft)
ts_segments.append(ts)
return ts_segments
def get_media_duration(obj):
'''
Returns a media objects duration in seconds. Assumes the object has a
`audio_file_field`
'''
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(obj)
code = \
"ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 {0}".format(
tmp_file)
data = commands.getstatusoutput(code)
return float(data[1])
def transcribe_segment(ts):
try:
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(ts.parent)
except Exception as e:
return "Error creating temporary environment. {0}".format(e)
tmp_seg_file = tmp_stor_dir + '/ts_{0}.wav'.format(ts.pk)
command = \
['ffmpeg', '-i', tmp_file,
'-ss', '{0:.2f}'.format(ts.start/100.0),
'-to', '{0:.2f}'.format(ts.end/100.0),
'-ar', '16000', '-ac', '1', # '-f', 's16le',
tmp_seg_file]
logger.debug("COMMAND: {0}".format(' '.join(command)))
p = Popen(command, stdin=PIPE, stdout=PIPE)
output, errors = p.communicate()
# result = transcribe_audio_sphinx(output)
result = transcribe_audio_sphinx(None,
continuous=True,
file_path=tmp_seg_file)
if result['success']:
# ts.text = parse_sphinx_transcription(result['transcription'])
ts.text = result['transcription'].strip()
ts.transcriber_log = result
# Get or create a source for the API
source, created = Source.objects.get_or_create(
source_name='Transcription API',
author="{0}".format(result['model_version']),
source_type='M',
source_url=result['API_URL'])
ts.source = source
ts.save()
else:
ts.transcriber_log = result
ts.save()
def slice_audio(aft, start, stop):
try:
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(aft)
except Exception as e:
return "Error creating temporary environment. {0}".format(e)
tmp_seg_file = tmp_stor_dir + '/sliced_{0}_{1}.wav'.format(start, stop)
command = \
['ffmpeg', '-i', tmp_file,
'-ss', '{0:.2f}'.format(start/100.0),
'-to', '{0:.2f}'.format(stop/100.0),
# '-ar', '16000', '-ac', '1', # '-f', 's16le',
tmp_seg_file]
logger.debug("COMMAND: {0}".format(' '.join(command)))
p = Popen(command, stdin=PIPE, stdout=PIPE)
output, errors = p.communicate()
return tmp_seg_file
def transcribe_segment(ts):
if not check_to_transcribe_segment(ts):
return 'Not transcribing segment. Likely segment too long.'
try:
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(ts.parent)
except Exception as e:
return "Error creating temporary environment. {0}".format(e)
tmp_seg_file = tmp_stor_dir + '/ts_{0}.wav'.format(ts.pk)
command = \
['ffmpeg', '-i', tmp_file,
'-ss', '{0:.2f}'.format(ts.start/100.0),
'-to', '{0:.2f}'.format(ts.end/100.0),
'-ar', '16000', '-ac', '1', # '-f', 's16le',
tmp_seg_file]
logger.debug("COMMAND: {0}".format(' '.join(command)))
p = Popen(command, stdin=PIPE, stdout=PIPE)
output, errors = p.communicate()
# result = transcribe_audio_sphinx(output)
result = transcribe_audio_sphinx(
None, continuous=True, file_path=tmp_seg_file)
if result['success']:
# ts.text = parse_sphinx_transcription(result['transcription'])
ts.text = result['transcription'].strip()
result['status'] = unicode(_('Done'))
ts.transcriber_log = result
# Get or create a source for the API
source, created = Source.objects.get_or_create(
source_name='Transcription API',
author="{0}".format(result['model_version']),
source_type='M',
source_url=result['API_URL'])
ts.source = source
if ts.text is '' or ts.text is ' ':
ts.no_speech_detected = True
ts.save()
else:
result['status'] = unicode(_('Error'))
if ts.transcriber_log:
if 'retry' in ts.transcriber_log.keys():
ts.transcriber_log.update(result)
ts.transcriber_log['retry'] = False
else:
ts.transcriber_log['retry'] = True
else:
ts.transcriber_log = result
ts.save()
os.remove(tmp_seg_file)
return result
def encode_audio(obj, test=False, codec='aac'):
codecs = {
'mp3': ['libmp3lame', 'mp3'],
'aac': ['aac', 'm4a', 44100, 1],
'wav': ['pcm_s16le', 'wav', 16000, 1]
}
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(obj)
# If a video doesn't have audio this will fail.
command = 'ffprobe -v quiet -show_entries stream -print_format json ' + \
tmp_file
data = ast.literal_eval(commands.getoutput(command))
streams = data['streams']
audio = False
for stream in streams:
if stream['codec_type'] in 'audio':
audio = True
if audio:
file_name = obj.get_file_name() + '_x'
extension = codecs[codec][1]
if codec in 'wav':
code = "ffmpeg -i {0} -vn -acodec {1} -ar {2} -ac {3} {4}/{5}.{6}".format(
tmp_file, codecs[codec][0], codecs[codec][2], codecs[codec][3],
tmp_stor_dir, file_name, extension)
elif codec in 'aac':
code = \
"ffmpeg -i {0} -vn -acodec {1} -ar {5} -ac {6} {7} {2}/{3}.{4}".format(
tmp_file, codecs[codec][0], tmp_stor_dir, file_name,
extension, codecs[codec][2], codecs[codec][3],
"-c:a aac -b:a 64k") # -profile:a aac_he
logger.debug('Running: ' + code)
data = commands.getstatusoutput(code)
logger.debug(data[1])
logger.debug(u'FILE FILENAME: \t{0}'.format(file_name))
if file_name is None:
file_name = 'audio'
if 'aac' in codec:
obj.audio_file_aac.save(
file_name + '.' + extension,
File(
open(tmp_stor_dir +
'/{0}.{1}'.format(file_name, extension))))
elif 'wav' in codec:
obj.audio_file_wav.save(
file_name + '.' + extension,
File(
open(tmp_stor_dir +
'/{0}.{1}'.format(file_name, extension))))
code = 'rm ' + tmp_stor_dir + '/{0}.{1}'.format(file_name, extension)
logger.debug('Running: ' + code)
data = commands.getstatusoutput(code)
logger.debug(data[1])
if not audio:
logger.debug('No audio stream found.')
return False
# Need a better way to check this!
data = commands.getstatusoutput('rm ' + tmp_file)
logger.debug('Removed tmp file %s' % (tmp_file))
data = commands.getstatusoutput('rm -r ' + tmp_stor_dir)
logger.debug('Removed tmp stor dir %s' % (tmp_stor_dir))
set_s3_content_deposition(obj)
return "Encoded {0}".format(obj)
def encode_audio(recording, test=False, codec='aac'):
'''
Encode audio into required formats. Also sets recording duration.
'''
if not isinstance(codec, list):
codec_list = [codec] # Backwards compatibility
else:
codec_list = codec
codecs = {
'mp3': ['libmp3lame', 'mp3'],
'aac': ['aac', 'm4a'],
'wav': ['pcm_s16le', 'wav', 16000, 1]
}
file_path, tmp_stor_dir, tmp_file, absolute_directory = \
prepare_temporary_environment(recording)
# If a video doesn't have audio this will fail.
command = 'ffprobe -v quiet -show_entries stream -print_format json ' + \
tmp_file
data = ast.literal_eval(commands.getoutput(command))
streams = data['streams']
audio = False
for stream in streams:
if stream['codec_type'] in 'audio':
audio = True
if audio:
file_name = recording.get_recording_file_name()
code = \
"ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 {0}".format(
tmp_file)
data = commands.getstatusoutput(code)
recording.duration = float(data[1])
for codec in codec_list:
if codec in 'wav':
file_name = file_name + '_16kHz'
extension = codecs[codec][1]
if codec in 'wav':
code = "ffmpeg -i {0} -vn -acodec {1} -ar {2} -ac {3} {4}/{5}.{6}".format(
tmp_file,
codecs[codec][0], codecs[codec][2], codecs[codec][3],
tmp_stor_dir, file_name, extension)
else:
code = "ffmpeg -i {0} -vn -acodec {1} {2}/{3}.{4}".format(
tmp_file, codecs[codec][0], tmp_stor_dir, file_name, extension)
logger.debug('Running: '+code)
data = commands.getstatusoutput(code)
logger.debug(data[1])
logger.debug(u'FILE FILENAME: \t{0}'.format(file_name))
if file_name is None:
file_name = 'audio'
if 'aac' in codec:
recording.audio_file_aac.save(
file_name+'.'+extension,
File(open(tmp_stor_dir+'/{0}.{1}'.format(
file_name, extension))))
elif 'wav' in codec:
recording.audio_file_wav.save(
file_name+'.'+extension,
File(open(tmp_stor_dir+'/{0}.{1}'.format(
file_name, extension))))
code = 'rm '+tmp_stor_dir+'/{0}.{1}'.format(file_name, extension)
logger.debug('Running: '+code)
data = commands.getstatusoutput(code)
logger.debug(data[1])
if not audio:
logger.debug('No audio stream found.')
return False
data = commands.getstatusoutput('rm ' + tmp_file)
logger.debug('Removed tmp file %s' % (tmp_file))
data = commands.getstatusoutput('rm -r ' + tmp_stor_dir)
logger.debug('Removed tmp stor dir %s' % (tmp_stor_dir))
set_s3_content_deposition(recording)
return "Encoded {0}".format(recording)
