def test_from_time(self):
time_obj = time(1, 2, 3, 4000)
self.assertEqual(WebVTTTime(1, 2, 3, 4), time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 5) >= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 3) <= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 0) != time_obj)
self.assertEqual(WebVTTTime(1, 2, 3, 4).to_time(), time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 5).to_time() >= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 3).to_time() <= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 0).to_time() != time_obj)
class TestSimpleTime(unittest.TestCase):
def setUp(self):
self.time = WebVTTTime()
def test_default_value(self):
self.assertEqual(self.time.ordinal, 0)
def test_micro_seconds(self):
self.time.milliseconds = 1
self.assertEqual(self.time.milliseconds, 1)
self.time.hours += 42
self.assertEqual(self.time.milliseconds, 1)
self.time.milliseconds += 1000
self.assertEqual(self.time.seconds, 1)
def test_seconds(self):
self.time.seconds = 1
self.assertEqual(self.time.seconds, 1)
self.time.hours += 42
self.assertEqual(self.time.seconds, 1)
self.time.seconds += 60
self.assertEqual(self.time.minutes, 1)
def test_minutes(self):
self.time.minutes = 1
self.assertEqual(self.time.minutes, 1)
self.time.hours += 42
self.assertEqual(self.time.minutes, 1)
self.time.minutes += 60
self.assertEqual(self.time.hours, 43)
def test_hours(self):
self.time.hours = 1
self.assertEqual(self.time.hours, 1)
self.time.minutes += 42
self.assertEqual(self.time.hours, 1)
def test_shifting(self):
self.time.shift(1, 1, 1, 1)
self.assertEqual(self.time, (1, 1, 1, 1))
def test_descriptor_from_class(self):
self.assertRaises(AttributeError, lambda: WebVTTTime.hours)
class TestSimpleTime(unittest.TestCase):
def setUp(self):
self.time = WebVTTTime()
def test_default_value(self):
self.assertEqual(self.time.ordinal, 0)
def test_micro_seconds(self):
self.time.milliseconds = 1
self.assertEqual(self.time.milliseconds, 1)
self.time.hours += 42
self.assertEqual(self.time.milliseconds, 1)
self.time.milliseconds += 1000
self.assertEqual(self.time.seconds, 1)
def test_seconds(self):
self.time.seconds = 1
self.assertEqual(self.time.seconds, 1)
self.time.hours += 42
self.assertEqual(self.time.seconds, 1)
self.time.seconds += 60
self.assertEqual(self.time.minutes, 1)
def test_minutes(self):
self.time.minutes = 1
self.assertEqual(self.time.minutes, 1)
self.time.hours += 42
self.assertEqual(self.time.minutes, 1)
self.time.minutes += 60
self.assertEqual(self.time.hours, 43)
def test_hours(self):
self.time.hours = 1
self.assertEqual(self.time.hours, 1)
self.time.minutes += 42
self.assertEqual(self.time.hours, 1)
def test_shifting(self):
self.time.shift(1, 1, 1, 1)
self.assertEqual(self.time, (1, 1, 1, 1))
def test_descriptor_from_class(self):
self.assertRaises(AttributeError, lambda: WebVTTTime.hours)
def test_from_tuple(self):
self.assertEqual((0, 0, 0, 0), WebVTTTime())
self.assertEqual((0, 0, 0, 1), WebVTTTime(milliseconds=1))
self.assertEqual((0, 0, 2, 0), WebVTTTime(seconds=2))
self.assertEqual((0, 3, 0, 0), WebVTTTime(minutes=3))
self.assertEqual((4, 0, 0, 0), WebVTTTime(hours=4))
self.assertEqual((1, 2, 3, 4), WebVTTTime(1, 2, 3, 4))
def test_from_dict(self):
self.assertEqual(dict(), WebVTTTime())
self.assertEqual(dict(milliseconds=1), WebVTTTime(milliseconds=1))
self.assertEqual(dict(seconds=2), WebVTTTime(seconds=2))
self.assertEqual(dict(minutes=3), WebVTTTime(minutes=3))
self.assertEqual(dict(hours=4), WebVTTTime(hours=4))
self.assertEqual(dict(hours=1, minutes=2, seconds=3, milliseconds=4),
WebVTTTime(1, 2, 3, 4))
class TestCoercing(TestCase):
def test_from_tuple(self):
self.assertEqual((0, 0, 0, 0), WebVTTTime())
self.assertEqual((0, 0, 0, 1), WebVTTTime(milliseconds=1))
self.assertEqual((0, 0, 2, 0), WebVTTTime(seconds=2))
self.assertEqual((0, 3, 0, 0), WebVTTTime(minutes=3))
self.assertEqual((4, 0, 0, 0), WebVTTTime(hours=4))
self.assertEqual((1, 2, 3, 4), WebVTTTime(1, 2, 3, 4))
def test_from_dict(self):
self.assertEqual(dict(), WebVTTTime())
self.assertEqual(dict(milliseconds=1), WebVTTTime(milliseconds=1))
self.assertEqual(dict(seconds=2), WebVTTTime(seconds=2))
self.assertEqual(dict(minutes=3), WebVTTTime(minutes=3))
self.assertEqual(dict(hours=4), WebVTTTime(hours=4))
self.assertEqual(dict(hours=1, minutes=2, seconds=3, milliseconds=4),
WebVTTTime(1, 2, 3, 4))
def test_from_time(self):
time_obj = time(1, 2, 3, 4000)
self.assertEqual(WebVTTTime(1, 2, 3, 4), time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 5) >= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 3) <= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 0) != time_obj)
self.assertEqual(WebVTTTime(1, 2, 3, 4).to_time(), time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 5).to_time() >= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 3).to_time() <= time_obj)
self.assertTrue(WebVTTTime(1, 2, 3, 0).to_time() != time_obj)
def test_from_ordinal(self):
self.assertEqual(WebVTTTime.from_ordinal(3600000), {'hours': 1})
self.assertEqual(WebVTTTime(1), 3600000)
def test_from_repr(self):
self.time = WebVTTTime()
self.assertEqual('WebVTTTime(0, 0, 0, 0)', self.time.__repr__())
self.time = WebVTTTime(1, 1, 1, 1)
self.assertEqual('WebVTTTime(1, 1, 1, 1)', self.time.__repr__())
def merge_subtitle(sub_a, sub_b, delta):
out = WebVTTFile()
intervals = [item.start.ordinal for item in sub_a]
intervals.extend([item.end.ordinal for item in sub_a])
intervals.extend([item.start.ordinal for item in sub_b])
intervals.extend([item.end.ordinal for item in sub_b])
intervals.sort()
j = k = 0
for i in xrange(1, len(intervals)):
start = WebVTTTime.from_ordinal(intervals[i - 1])
end = WebVTTTime.from_ordinal(intervals[i])
if (end-start) > delta:
text_a, j = find_subtitle(sub_a, start, end, j)
text_b, k = find_subtitle(sub_b, start, end, k)
text = join_lines(text_a, text_b)
if len(text) > 0:
item = WebVTTItem(0, start, end, text)
out.append(item)
out.clean_indexes()
return out
def merge_subtitle(sub_a, sub_b, delta):
out = WebVTTFile()
intervals = [item.start.ordinal for item in sub_a]
intervals.extend([item.end.ordinal for item in sub_a])
intervals.extend([item.start.ordinal for item in sub_b])
intervals.extend([item.end.ordinal for item in sub_b])
intervals.sort()
j = k = 0
for i in xrange(1, len(intervals)):
start = WebVTTTime.from_ordinal(intervals[i - 1])
end = WebVTTTime.from_ordinal(intervals[i])
if (end - start) > delta:
text_a, j = find_subtitle(sub_a, start, end, j)
text_b, k = find_subtitle(sub_b, start, end, k)
text = join_lines(text_a, text_b)
if len(text) > 0:
item = WebVTTItem(0, start, end, text)
out.append(item)
out.clean_indexes()
return out
class TestSimpleTime(TestCase):
def setUp(self):
self.time = WebVTTTime()
def test_default_value(self):
self.assertEqual(self.time.ordinal, 0)
def test_micro_seconds(self):
self.time.milliseconds = 1
self.assertEqual(self.time.milliseconds, 1)
self.time.hours += 42
self.assertEqual(self.time.milliseconds, 1)
self.time.milliseconds += 1000
self.assertEqual(self.time.seconds, 1)
def test_seconds(self):
self.time.seconds = 1
self.assertEqual(self.time.seconds, 1)
self.time.hours += 42
self.assertEqual(self.time.seconds, 1)
self.time.seconds += 60
self.assertEqual(self.time.minutes, 1)
def test_minutes(self):
self.time.minutes = 1
self.assertEqual(self.time.minutes, 1)
self.time.hours += 42
self.assertEqual(self.time.minutes, 1)
self.time.minutes += 60
self.assertEqual(self.time.hours, 43)
def test_hours(self):
self.time.hours = 1
self.assertEqual(self.time.hours, 1)
self.time.minutes += 42
self.assertEqual(self.time.hours, 1)
def test_shifting_forward(self):
self.time.shift(1, 1, 1, 1)
self.assertEqual(self.time, (1, 1, 1, 1))
def test_shifting_backwards(self):
self.time.shift(-1, -1, -1, -1)
self.assertEqual(self.time, (-2, 58, 58, 999))
self.time = WebVTTTime(1, 2, 3, 4)
self.time.shift(-1, -1, -1, -1)
self.assertEqual(self.time, (0, 1, 2, 3))
def test_descriptor_from_class(self):
self.assertRaises(AttributeError, lambda: WebVTTTime.hours)
def test_negative_serialization(self):
self.assertEqual('00:00:00.000', str(WebVTTTime(-1, 2, 3, 4)))
def test_serialization(self):
for time_string, time_items in self.KNOWN_VALUES:
self.assertEqual(time_string, str(WebVTTTime(*time_items)))
def test_parsing(self):
for time_string, time_items in self.KNOWN_VALUES:
self.assertEqual(time_string, WebVTTTime(*time_items))
def setUp(self):
self.time = WebVTTTime()
def test_mul(self):
self.assertEqual(self.time * 2, WebVTTTime(2, 4, 6, 8))
self.assertEqual(self.time * 0.5, (0, 31, 1, 502))
def test_from_repr(self):
self.time = WebVTTTime()
self.assertEqual('WebVTTTime(0, 0, 0, 0)', self.time.__repr__())
self.time = WebVTTTime(1, 1, 1, 1)
self.assertEqual('WebVTTTime(1, 1, 1, 1)', self.time.__repr__())
def test_invalid_int(self):
random_long = int(choice(list(range(0, 10000000))))
# String
self.assertRaises(ValueError, lambda: WebVTTTime.parse_int('test'))
# Binary
self.assertRaises(ValueError, lambda: WebVTTTime.parse_int(bin(42)))
# Char
self.assertRaises(ValueError, lambda: WebVTTTime.parse_int('t'))
# None
self.assertRaises(TypeError, WebVTTTime.parse_int)
# None
self.assertRaises(TypeError, WebVTTTime.parse_int(None))
# List
self.assertRaises(TypeError, WebVTTTime.parse_int(list(range(10))))
# Tuple
self.assertRaises(TypeError, WebVTTTime.parse_int((1, 1)))
# Boolean
self.assertRaises(TypeError, WebVTTTime.parse_int(True))
# Float
self.assertRaises(TypeError, WebVTTTime.parse_int(uniform(1, 100)))
# Complex
self.assertRaises(TypeError, WebVTTTime.parse_int(1j))
# Long
self.assertRaises(TypeError, WebVTTTime.parse_int(random_long))
# Dictionary
self.assertRaises(TypeError,
WebVTTTime.parse_int({
'Test1': 1,
'Test0': 0
}))
def test_from_ordinal(self):
self.assertEqual(WebVTTTime.from_ordinal(3600000), {'hours': 1})
self.assertEqual(WebVTTTime(1), 3600000)
def test_shifting_backwards(self):
self.time.shift(-1, -1, -1, -1)
self.assertEqual(self.time, (-2, 58, 58, 999))
self.time = WebVTTTime(1, 2, 3, 4)
self.time.shift(-1, -1, -1, -1)
self.assertEqual(self.time, (0, 1, 2, 3))
def main(options):
# Ensure ffmpeg is around
if not run_ffmpeg(['-version']):
log.error(
"ffmpeg needs to be available to strip audio from the video file.")
exit(1)
with NamedTemporaryFile(delete=True) as vid_file:
log.info("Downloading %s - this might take a while." % options.vid_url)
response = get(options.vid_url, stream=True)
total_length = response.headers.get("content-length")
if total_length is None: # no content length header
log.info("Unknown length - can't predict how long this will take.")
f.write(response.content)
else:
bar = ProgressBar(max_value=int(total_length))
dl = 0
for data in response.iter_content(chunk_size=DOWNLOAD_CHUNK_SIZE):
dl += len(data)
vid_file.write(data)
vid_file.flush()
bar.update(dl)
log.info("Download done. Stripping audio.")
(wav_file, wav_file_name) = mkstemp('.wav')
result = run_ffmpeg([
"-y", "-i", vid_file.name, "-vn", "-acodec", "pcm_s16le", "-ar",
"16000", "-ac", "1", wav_file_name
])
if not result:
close(wav_file)
log.error("ffmpeg failed. Bailing.")
exit(1)
fs, audio = wav.read(wav_file_name)
close(wav_file)
log.info("Will write VTT to %s" % options.output)
# Make sure the WAV is to code...
log.info("Loading up WAV file...")
if fs != 16000:
log.error("Only 16000hz WAV files are usable.")
exit(1)
total_samples = len(audio)
duration_hours, duration_minutes, duration_seconds = sample_index_to_time(
len(audio))
log.info("Approximate duration: %d:%02d:%02d" %
(duration_hours, duration_minutes, duration_seconds))
# Let's load up DeepSpeech and get it ready.
log.info("Loading pre-trained DeepSpeech model...")
root_model_dir = path.join(options.deepspeech_model_dir, MODEL_DIR)
model = path.join(root_model_dir, MODEL_FILE)
alphabet = path.join(root_model_dir, MODEL_ALPHABET)
lang_model = path.join(root_model_dir, MODEL_LANG_MODEL)
trie = path.join(root_model_dir, MODEL_TRIE)
deepspeech = Model(model, N_FEATURES, N_CONTEXT, alphabet, BEAM_WIDTH)
log.info("Done loading model.")
log.info("Loading language model...")
deepspeech.enableDecoderWithLM(alphabet, lang_model, trie, LM_WEIGHT,
WORD_COUNT_WEIGHT, VALID_WORD_COUNT_WEIGHT)
log.info("Done loading model.")
playhead = 0
out = WebVTTFile()
bar = ProgressBar(max_value=total_samples)
while playhead < (total_samples - 1):
end_point = min(playhead + AUDIO_SEGMENT_SAMPLES, (total_samples - 1))
segment = audio[playhead:end_point]
inference = deepspeech.stt(segment, fs)
log.debug("Inferred: %s" % inference)
start_hours, start_minutes, start_seconds = sample_index_to_time(
playhead)
playhead = end_point
end_hours, end_minutes, end_seconds = sample_index_to_time(playhead)
if not inference or inference == "ah":
continue
for search, replace in INFERENCE_REPLACEMENTS.iteritems():
inference = sub(r"\b" + search + r"\b", replace, inference)
inference = fill(inference, width=MAX_CAPTION_WIDTH)
start = WebVTTTime(start_hours, start_minutes, start_seconds)
end = WebVTTTime(end_hours, end_minutes, end_seconds)
item = WebVTTItem(0, start, end, inference)
out.append(item)
bar.update(playhead)
out.save(options.output, encoding="utf-8")
out.clean_indexes()
out.save(options.output, encoding="utf-8")
def test_negative_serialization(self):
self.assertEqual('00:00:00.000', str(WebVTTTime(-1, 2, 3, 4)))
self.assertEqual('00:00:00.000', str(WebVTTTime(-maxsize, 2, 3, 4)))
self.assertEqual('00:00:00.000', str(WebVTTTime(0, -2, 3, 4)))
self.assertEqual('00:00:00.000', str(WebVTTTime(0, 0, -3, 4)))
self.assertEqual('00:00:00.000', str(WebVTTTime(0, 0, 0, -4)))
print " --delta=<milliseconds> default: 500"
print " -e <encoding> Encoding of input and output files."
print " --encoding=<encoding> default: utf_8"
def main():
try:
opts, args = getopt.getopt(sys.argv[1:], 'hd:e:',
["help", "encoding=", "delta="])
except getopt.GetoptError, err:
print str(err)
usage()
sys.exit(2)
#Settings default values
delta = WebVTTTime(milliseconds=500)
encoding = "utf_8"
#-
if len(args) <> 3:
usage()
sys.exit(2)
for o, a in opts:
if o in ("-d", "--delta"):
delta = WebVTTTime(milliseconds=int(a))
elif o in ("-e", "--encoding"):
encoding = a
elif o in ("-h", "--help"):
usage()
sys.exit()
def setUp(self):
self.time = WebVTTTime()
def setUp(self):
self.time = WebVTTTime(1, 2, 3, 4)
def test_max_values(self):
self.assertEqual('99:59:59.999', str(WebVTTTime(99, 59, 59, 999)))
self.assertEqual('100:40:39.999', str(WebVTTTime(99, 99, 99, 999)))
def test_from_ordinal(self):
self.assertEqual(WebVTTTime.from_ordinal(3600000), {'hours': 1})
self.assertEqual(WebVTTTime(1), 3600000)
