def test_wav_multiple_channels(self, dtype, sample_rate, num_channels):
"""`sox_io_backend.info` can check wav file with channels more than 2 correctly"""
duration = 1
path = self.get_temp_path('data.wav')
data = get_wav_data(dtype,
num_channels,
normalize=False,
num_frames=duration * sample_rate)
save_wav(path, data, sample_rate)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == sox_utils.get_bit_depth(dtype)
assert info.encoding == get_encoding('wav', dtype)
def test_htk(self):
"""`sox_io_backend.info` can check HTK file correctly"""
duration = 1
num_channels = 1
sample_rate = 8000
path = self.get_temp_path('data.htk')
sox_utils.gen_audio_file(
path, sample_rate=sample_rate, num_channels=num_channels,
bit_depth=16, duration=duration)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == 16
assert info.encoding == "PCM_S"
def test_gsm(self):
"""`sox_io_backend.info` can check gsm file correctly"""
duration = 1
num_channels = 1
sample_rate = 8000
path = self.get_temp_path('data.gsm')
sox_utils.gen_audio_file(path,
sample_rate=sample_rate,
num_channels=num_channels,
duration=duration)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_channels == num_channels
assert info.bits_per_sample == 0
assert info.encoding == "GSM"
def test_vorbis(self, sample_rate, num_channels, quality_level):
"""`sox_io_backend.info` can check vorbis file correctly"""
duration = 1
path = self.get_temp_path('data.vorbis')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
compression=quality_level,
duration=duration,
)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
def test_sphere(self, sample_rate, num_channels, bits_per_sample):
"""`sox_io_backend.info` can check sph file correctly"""
duration = 1
path = self.get_temp_path('data.sph')
sox_utils.gen_audio_file(path,
sample_rate,
num_channels,
duration=duration,
bit_depth=bits_per_sample)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == bits_per_sample
assert info.encoding == "PCM_S"
def test_flac(self, sample_rate, num_channels, compression_level):
"""`sox_io_backend.info` can check flac file correctly"""
duration = 1
path = self.get_temp_path('data.flac')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
compression=compression_level,
duration=duration,
)
info = sox_io_backend.info(path)
assert info.get_sample_rate() == sample_rate
assert info.get_num_frames() == sample_rate * duration
assert info.get_num_channels() == num_channels
def test_mp3(self, sample_rate, num_channels, bit_rate):
"""`sox_io_backend.info` can check mp3 file correctly"""
duration = 1
path = self.get_temp_path('data.mp3')
sox_utils.gen_audio_file(
path, sample_rate, num_channels,
compression=bit_rate, duration=duration,
)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
# mp3 does not preserve the number of samples
# assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == 0 # bit_per_sample is irrelevant for compressed formats
assert info.encoding == "MP3"
def test_mp3(self, sample_rate, num_channels, bit_rate):
"""`sox_io_backend.info` can check mp3 file correctly"""
duration = 1
path = self.get_temp_path('data.mp3')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
compression=bit_rate,
duration=duration,
)
info = sox_io_backend.info(path)
assert info.get_sample_rate() == sample_rate
# mp3 does not preserve the number of samples
# assert info.get_num_frames() == sample_rate * duration
assert info.get_num_channels() == num_channels
def test_wav_multiple_channels(self, dtype, sample_rate, num_channels):
"""`sox_io_backend.info` can check wav file with channels more than 2 correctly"""
duration = 1
path = self.get_temp_path(f'{dtype}_{sample_rate}_{num_channels}.wav')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
bit_depth=sox_utils.get_bit_depth(dtype),
encoding=sox_utils.get_encoding(dtype),
duration=duration,
)
info = sox_io_backend.info(path)
assert info.get_sample_rate() == sample_rate
assert info.get_num_samples() == sample_rate * duration
assert info.get_num_channels() == num_channels
def test_vorbis(self, sample_rate, num_channels, quality_level):
"""`sox_io_backend.info` can check vorbis file correctly"""
duration = 1
path = self.get_temp_path(
f'{sample_rate}_{num_channels}_{quality_level}.vorbis')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
compression=quality_level,
duration=duration,
)
info = sox_io_backend.info(path)
assert info.get_sample_rate() == sample_rate
assert info.get_num_samples() == sample_rate * duration
assert info.get_num_channels() == num_channels
def test_amb(self, dtype, sample_rate, num_channels):
"""`sox_io_backend.info` can check amb file correctly"""
duration = 1
path = self.get_temp_path('data.amb')
bits_per_sample = sox_utils.get_bit_depth(dtype)
sox_utils.gen_audio_file(path,
sample_rate,
num_channels,
bit_depth=bits_per_sample,
duration=duration)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == bits_per_sample
assert info.encoding == get_encoding("amb", dtype)
def test_alaw(self):
"""`sox_io_backend.info` can check alaw file correctly"""
duration = 1
num_channels = 1
sample_rate = 8000
path = self.get_temp_path('data.wav')
sox_utils.gen_audio_file(
path, sample_rate=sample_rate, num_channels=num_channels,
bit_depth=8, encoding='a-law',
duration=duration)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == 8
assert info.encoding == "ALAW"
def test_mp3(self):
"""Providing `format` allows to read mp3 without extension
libsox does not check header for mp3
https://github.com/pytorch/audio/issues/1040
The file was generated with the following command
ffmpeg -f lavfi -i "sine=frequency=1000:duration=5" -ar 16000 -f mp3 test_noext
"""
path = get_asset_path("mp3_without_ext")
sinfo = sox_io_backend.info(path, format="mp3")
assert sinfo.sample_rate == 16000
assert sinfo.num_frames == 81216
assert sinfo.num_channels == 1
assert sinfo.bits_per_sample == 0 # bit_per_sample is irrelevant for compressed formats
assert sinfo.encoding == "MP3"
def test_vorbis(self, sample_rate, num_channels, quality_level):
"""`sox_io_backend.info` can check vorbis file correctly"""
duration = 1
path = self.get_temp_path('data.vorbis')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
compression=quality_level,
duration=duration,
)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == 0 # bit_per_sample is irrelevant for compressed formats
assert info.encoding == "VORBIS"
def test_flac(self, sample_rate, num_channels, compression_level):
"""`sox_io_backend.info` can check flac file correctly"""
duration = 1
path = self.get_temp_path('data.flac')
sox_utils.gen_audio_file(
path,
sample_rate,
num_channels,
compression=compression_level,
duration=duration,
)
info = sox_io_backend.info(path)
assert info.sample_rate == sample_rate
assert info.num_frames == sample_rate * duration
assert info.num_channels == num_channels
assert info.bits_per_sample == 24 # FLAC standard
assert info.encoding == "FLAC"
def _query_fileobj(self,
ext,
dtype,
sample_rate,
num_channels,
num_frames,
*,
comments=None):
path = self._gen_file(ext,
dtype,
sample_rate,
num_channels,
num_frames,
comments=comments)
format_ = ext if ext in ['mp3'] else None
with open(path, 'rb') as fileobj:
return sox_io_backend.info(fileobj, format_)
def py_info_func(filepath: str) -> torch.classes.torchaudio.SignalInfo:
return sox_io_backend.info(filepath)
def _query_bytesio(self, ext, dtype, sample_rate, num_channels, num_frames):
path = self._gen_file(ext, dtype, sample_rate, num_channels, num_frames)
format_ = ext if ext in ['mp3'] else None
with open(path, 'rb') as file_:
fileobj = io.BytesIO(file_.read())
return sox_io_backend.info(fileobj, format_)
