def split_with_vad_wav(
wav_path: Path,
out_dir: Path,
min_dur: float,
max_dur: float,
max_silence: float,
strict_min_dur: bool,
shift: float = 0,
) -> None:
assert Path(wav_path).suffix == ".wav"
audio_region = AudioRegion.load(str(wav_path))
out_dir = Path(out_dir)
regions = audio_region.split(
min_dur=min_dur,
max_dur=max_dur,
max_silence=max_silence,
strict_min_dur=strict_min_dur,
)
waveform, sr = sf.read(wav_path, dtype="float32")
out = []
for i, r in enumerate(regions):
start = int(r._meta.start * sr)
end = int(r._meta.end * sr)
path_seg = out_dir / f"{out_dir.stem}_{i}.flac"
path_timestamp = get_path_timestamp(path_seg, ".vad.timestamp")
save_timestamp(path_timestamp, r._meta.start + shift, r._meta.end + shift)
sf.write(
str(path_seg), waveform[start:end], sr, subtype="PCM_16", format="FLAC"
)
out.append(path_seg)
return out
def test_StreamSaverWorker_wav(self):
with TemporaryDirectory() as tmpdir:
expected_filename = os.path.join(tmpdir, "output.wav")
saver = StreamSaverWorker(self.reader, expected_filename)
saver.start()
tokenizer = TokenizerWorker(saver)
tokenizer.start_all()
tokenizer.join()
saver.join()
output_filename = saver.save_stream()
region = AudioRegion.load(
"tests/data/test_split_10HZ_mono.raw", sr=10, sw=2, ch=1
)
expected_region = AudioRegion.load(output_filename)
self.assertEqual(output_filename, expected_filename)
self.assertEqual(region, expected_region)
self.assertEqual(saver.data, bytes(expected_region))
def test_StreamSaverWorker_encode_audio(self):
with TemporaryDirectory() as tmpdir:
with patch("auditok.workers._run_subprocess") as patch_rsp:
patch_rsp.return_value = (1, None, None)
expected_filename = os.path.join(tmpdir, "output.ogg")
tmp_expected_filename = expected_filename + ".wav"
saver = StreamSaverWorker(self.reader, expected_filename)
saver.start()
tokenizer = TokenizerWorker(saver)
tokenizer.start_all()
tokenizer.join()
saver.join()
with self.assertRaises(AudioEncodingWarning) as rt_warn:
saver.save_stream()
warn_msg = "Couldn't save audio data in the desired format "
warn_msg += "'ogg'. Either none of 'ffmpeg', 'avconv' or 'sox' "
warn_msg += "is installed or this format is not recognized.\n"
warn_msg += "Audio file was saved as '{}'"
self.assertEqual(
warn_msg.format(tmp_expected_filename), str(rt_warn.exception)
)
ffmpef_avconv = [
"-y",
"-f",
"wav",
"-i",
tmp_expected_filename,
"-f",
"ogg",
expected_filename,
]
expected_calls = [
call(["ffmpeg"] + ffmpef_avconv),
call(["avconv"] + ffmpef_avconv),
call(
[
"sox",
"-t",
"wav",
tmp_expected_filename,
expected_filename,
]
),
]
self.assertEqual(patch_rsp.mock_calls, expected_calls)
region = AudioRegion.load(
"tests/data/test_split_10HZ_mono.raw", sr=10, sw=2, ch=1
)
self.assertTrue(saver._exported)
self.assertEqual(saver.data, bytes(region))