def testMetaData(self):
source = pyvideo.load("test_media/test_video.mp4")
self.assertAlmostEqual(source.duration, 10.041667)
video_format = source.video_format
self.assertEqual(video_format.height, 480)
self.assertEqual(video_format.width, 854)
audio_format = source.audio_format
self.assertEqual(audio_format.channels, 2)
self.assertEqual(audio_format.sample_size, 16)
self.assertEqual(audio_format.sample_rate, 48000)
def get_audio_from_file(filename):
"""
Extracts audio data from video file in mono
Returns tuple of (sample_rate, numpy samples array)
"""
import pyvideo
source = pyvideo.load(filename, skip_video=True)
video_ts = source.get_next_video_timestamp()
num_samples = source.duration * source.audio_format.sample_rate * source.audio_format.channels
if source.audio_format.sample_size == 16:
file_data = numpy.zeros((num_samples,), dtype=numpy.int16)
else:
assert source.audio_format.sample_size == 8
file_data = numpy.zeros((num_samples,), dtype=numpy.int8)
counter = 0
last = 0
while True:
audio_data = source.get_audio_data()
if audio_data is None:
break
# Consume video frames which prevents memory leakage
while video_ts < audio_data.timestamp:
source.get_next_video_frame()
video_ts = source.get_next_video_timestamp()
if video_ts is None:
break
percent = int(video_ts * 100 / source.duration)
if percent != last:
last = percent
data = numpy.fromstring(audio_data.get_string_data(), dtype=file_data.dtype)
if counter + data.shape[0] < file_data.shape[0]:
file_data[counter:counter + data.shape[0]] = data
counter += len(data)
else:
break
# Reshape to match channels
assert source.audio_format.channels == 2 or source.audio_format.channels == 1
if source.audio_format.channels == 2:
if len(file_data) % 2 == 1: file_data = numpy.resize(file_data, len(file_data) + 1) # Pad with zero if length is odd
file_data = numpy.reshape(file_data, (-1, source.audio_format.channels))
# Flatten channels to mono if audio is stereo
file_data = (file_data[:,0] + file_data[:, 1]) / 2
return file_data, source.audio_format.sample_rate
def load(self, keyframes_only=False):
try:
self.source = pyvideo.load(self.filepath, keyframes_only=keyframes_only)
except MediaException as e:
raise VideoLoadException("Unknown video format.\n%s" % (e,))
except IOError as e:
raise VideoLoadException("Could not load video file.\n%s" % (e,))
self._determine_fps()
self.info = VideoInfo(self.source.video_format.width,
self.source.video_format.height,
self._fps,
self.source.audio_format.channels,
self.source.audio_format.sample_rate,
self.source.duration)
def testFullDecode(self):
source = pyvideo.load("test_media/test_media.webm")
self.assertIsNotNone(source.audio_format)
self.assertIsNotNone(source.video_format)
timestamp = source.get_next_video_timestamp()
self.assertIsNotNone(timestamp)
while timestamp is not None:
frame = source.get_next_video_frame()
timestamp = source.get_next_video_timestamp()
# Decode audio data until the next frame
while True:
audio_data = source.get_audio_data()
if audio_data is None or audio_data.timestamp > timestamp:
break
