Features/cli

pyproject.toml

@@ -1,9 +1,12 @@
 [tool.poetry]
 name = "pytsmod"
 version = "0.2.0"
-description = ""
+description = "An open-source Python library for audio time-scale modification."
 authors = ["Sangeon Yong <koragon2@kaist.ac.kr>"]
 
+license = "GPL-3.0"
+readme = "README.md"
+
 [tool.poetry.dependencies]
 python = "^3.6"
 numpy = "^1.16.0"
@@ -15,8 +18,8 @@ librosa = "^0.8"
 pytest = "^5.2"
 flake8 = "^3.8.3"
 
-# [tool.poetry.scripts]
-# tsmod = 'pytsmod.console:run'
+[tool.poetry.scripts]
+tsmod = 'pytsmod.console:run'
 
 [build-system]
 requires = ["poetry>=0.12"]

pytsmod/console.py

@@ -0,0 +1,100 @@
+import sys
+sys.path.append('./')
+
+from pytsmod import ola, wsola
+from pytsmod import phase_vocoder as pv
+from pytsmod import phase_vocoder_int as pv_int
+import argparse
+import soundfile as sf
+
+
+def run():
+    parser = argparse.ArgumentParser(description='Processing time-scale modification for given audio file.')
+    parser.add_argument('algorithm', nargs='?', choices=['ola', 'wsola', 'pv', 'pv_int', 'hp'])
+    # parser.add_argument('--help', action='store_true')
+
+    args, sub_args = parser.parse_known_args()
+
+    if args.algorithm == 'ola':
+        parser = argparse.ArgumentParser()
+        parser.add_argument('input_file', type=str)
+        parser.add_argument('output_file', type=str)
+        parser.add_argument('alpha', type=float)
+        parser.add_argument('--win_type', '-wt', default='hann', type=str)
+        parser.add_argument('--win_size', '-ws', default=1024, type=int)
+        parser.add_argument('--syn_hop_size', '-sh', default=512, type=int)
+
+        params = parser.parse_args(sub_args)
+
+        x, sr = sf.read(params.input_file)
+
+        y = ola(x, params.alpha, win_type=params.win_type,
+                win_size=params.win_size, syn_hop_size=params.syn_hop_size)
+    elif args.algorithm == 'wsola':
+        parser = argparse.ArgumentParser()
+        parser.add_argument('input_file', type=str)
+        parser.add_argument('output_file', type=str)
+        parser.add_argument('alpha', type=float)
+        parser.add_argument('--win_type', '-wt', default='hann', type=str)
+        parser.add_argument('--win_size', '-ws', default=1024, type=int)
+        parser.add_argument('--syn_hop_size', '-sh', default=512, type=int)
+        parser.add_argument('--tolerance', '-t', default=512, type=int)
+
+        params = parser.parse_args(sub_args)
+
+        x, sr = sf.read(params.input_file)
+
+        y = wsola(x, params.alpha, win_type=params.win_type,
+                  win_size=params.win_size, syn_hop_size=params.syn_hop_size,
+                  tolerance=params.tolerance)
+    elif args.algorithm == 'pv':
+        parser = argparse.ArgumentParser()
+        parser.add_argument('input_file', type=str)
+        parser.add_argument('output_file', type=str)
+        parser.add_argument('alpha', type=float)
+        parser.add_argument('--win_type', '-wt', default='sin', type=str)
+        parser.add_argument('--win_size', '-ws', default=2048, type=int)
+        parser.add_argument('--syn_hop_size', '-sh', default=512, type=int)
+        parser.add_argument('--zero_pad', '-z', default=0, type=int)
+        parser.add_argument('--restore_energy', '-e', action='store_true')
+        parser.add_argument('--fft_shift', '-fs', action='store_true')
+        parser.add_argument('--phase_lock', '-pl', action='store_true')
+
+        params = parser.parse_args(sub_args)
+
+        x, sr = sf.read(params.input_file)
+
+        y = pv(x, params.alpha, win_type=params.win_type,
+               win_size=params.win_size, syn_hop_size=params.syn_hop_size,
+               zero_pad=params.zero_pad, restore_energy=params.restore_energy,
+               fft_shift=params.fft_shift, phase_lock=params.phase_lock)
+    elif args.algorithm == 'pv_int':
+        parser = argparse.ArgumentParser()
+        parser.add_argument('input_file', type=str)
+        parser.add_argument('output_file', type=str)
+        parser.add_argument('alpha', type=int)
+        parser.add_argument('--win_type', '-wt', default='hann', type=str)
+        parser.add_argument('--win_size', '-ws', default=2048, type=int)
+        parser.add_argument('--syn_hop_size', '-sh', default=512, type=int)
+        parser.add_argument('--zero_pad', '-z', default=None, type=int)
+        parser.add_argument('--restore_energy', '-e', action='store_true')
+        parser.add_argument('--fft_shift', '-fs', action='store_true')
+
+        params = parser.parse_args(sub_args)
+        print(params.zero_pad)
+
+        x, sr = sf.read(params.input_file)
+
+        y = pv_int(x, params.alpha, win_type=params.win_type,
+                   win_size=params.win_size, syn_hop_size=params.syn_hop_size,
+                   zero_pad=params.zero_pad,
+                   restore_energy=params.restore_energy,
+                   fft_shift=params.fft_shift)
+    # elif args.algorithm == 'hp':
+    #     pass
+
+    sf.write(params.output_file, y, sr)
+
+
+if __name__ == '__main__':
+    run()

pytsmod/pvtsm.py

@@ -172,7 +172,7 @@ def phase_vocoder_int(x, s, win_type='hann', win_size=2048, syn_hop_size=512,
 
         y[c, :] = y_chan
 
-    return y
+    return y.squeeze()
 
 
 def _find_peaks(spec):

tests/test_console.py

@@ -0,0 +1,155 @@
+import pytest
+from pytsmod import ola, wsola
+from pytsmod import phase_vocoder as pv
+from pytsmod import phase_vocoder_int as pv_int
+import soundfile as sf
+import numpy as np
+import os
+from subprocess import call
+
+
+@pytest.mark.parametrize('algorithm', ['ola', 'wsola', 'pv', 'pv_int'])
+def test_console_default_params(algorithm):
+    test_file = 'tests/data/castanetsviolin.wav'
+    alpha = 2
+    x, sr = sf.read(test_file)
+    y = globals()[algorithm](x, alpha)
+
+    cmd = ['python', 'pytsmod/console.py', algorithm,
+           test_file, 'temp_cli.wav', str(alpha)]
+    if algorithm == 'pv_int':
+        cmd.append('-fs')
+    call(cmd)
+
+    sf.write('temp.wav', y, sr)
+    y_, _ = sf.read('temp.wav')
+
+    y_cli, _ = sf.read('temp_cli.wav')
+
+    os.remove('temp.wav')
+    os.remove('temp_cli.wav')
+
+    assert np.allclose(y_, y_cli)
+
+
+@pytest.mark.parametrize('alpha', [1.25])
+@pytest.mark.parametrize('win_type', ['sin'])
+@pytest.mark.parametrize('win_size', [512])
+@pytest.mark.parametrize('syn_hop_size', [256])
+def test_console_ola(alpha, win_type, win_size, syn_hop_size):
+    test_file = 'tests/data/castanetsviolin.wav'
+    x, sr = sf.read(test_file)
+    y = ola(x, alpha, win_type=win_type, win_size=win_size,
+            syn_hop_size=syn_hop_size)
+
+    cmd = ['python', 'pytsmod/console.py', 'ola',
+           test_file, 'temp_cli.wav', str(alpha),
+           '-wt', win_type, '-ws', str(win_size),
+           '-sh', str(syn_hop_size)]
+    call(cmd)
+
+    sf.write('temp.wav', y, sr)
+    y_, _ = sf.read('temp.wav')
+
+    y_cli, _ = sf.read('temp_cli.wav')
+
+    os.remove('temp.wav')
+    os.remove('temp_cli.wav')
+
+    assert np.allclose(y_, y_cli)
+
+
+@pytest.mark.parametrize('alpha', [1.25])
+@pytest.mark.parametrize('win_type', ['sin'])
+@pytest.mark.parametrize('win_size', [512])
+@pytest.mark.parametrize('syn_hop_size', [256])
+@pytest.mark.parametrize('tolerance', [256])
+def test_console_wsola(alpha, win_type, win_size, syn_hop_size, tolerance):
+    test_file = 'tests/data/castanetsviolin.wav'
+    x, sr = sf.read(test_file)
+    y = wsola(x, alpha, win_type=win_type, win_size=win_size,
+              syn_hop_size=syn_hop_size, tolerance=tolerance)
+
+    cmd = ['python', 'pytsmod/console.py', 'wsola',
+           test_file, 'temp_cli.wav', str(alpha),
+           '-wt', win_type, '-ws', str(win_size),
+           '-sh', str(syn_hop_size), '-t', str(tolerance)]
+    call(cmd)
+
+    sf.write('temp.wav', y, sr)
+    y_, _ = sf.read('temp.wav')
+
+    y_cli, _ = sf.read('temp_cli.wav')
+
+    os.remove('temp.wav')
+    os.remove('temp_cli.wav')
+
+    assert np.allclose(y_, y_cli)
+
+
+@pytest.mark.parametrize('alpha', [1.25])
+@pytest.mark.parametrize('win_type', ['hann'])
+@pytest.mark.parametrize('win_size', [1024])
+@pytest.mark.parametrize('syn_hop_size', [256])
+@pytest.mark.parametrize('zero_pad', [256])
+@pytest.mark.parametrize('restore_energy', [True])
+@pytest.mark.parametrize('fft_shift', [True])
+@pytest.mark.parametrize('phase_lock', [True])
+def test_console_pv(alpha, win_type, win_size, syn_hop_size, zero_pad,
+                    restore_energy, fft_shift, phase_lock):
+    test_file = 'tests/data/castanetsviolin.wav'
+    x, sr = sf.read(test_file)
+    y = pv(x, alpha, win_type=win_type, win_size=win_size,
+           syn_hop_size=syn_hop_size, zero_pad=zero_pad,
+           restore_energy=restore_energy, fft_shift=fft_shift,
+           phase_lock=phase_lock)
+
+    cmd = ['python', 'pytsmod/console.py', 'pv',
+           test_file, 'temp_cli.wav', str(alpha),
+           '-wt', win_type, '-ws', str(win_size),
+           '-sh', str(syn_hop_size), '-z', str(zero_pad),
+           '-e', '-fs', '-pl']
+    call(cmd)
+
+    sf.write('temp.wav', y, sr)
+    y_, _ = sf.read('temp.wav')
+
+    y_cli, _ = sf.read('temp_cli.wav')
+
+    os.remove('temp.wav')
+    os.remove('temp_cli.wav')
+
+    assert np.allclose(y_, y_cli)
+
+
+@pytest.mark.parametrize('alpha', [2])
+@pytest.mark.parametrize('win_type', ['sin'])
+@pytest.mark.parametrize('win_size', [1024])
+@pytest.mark.parametrize('syn_hop_size', [256])
+@pytest.mark.parametrize('zero_pad', [256])
+@pytest.mark.parametrize('restore_energy', [True])
+@pytest.mark.parametrize('fft_shift', [False])
+def test_console_pv_int(alpha, win_type, win_size, syn_hop_size, zero_pad,
+                        restore_energy, fft_shift):
+    test_file = 'tests/data/castanetsviolin.wav'
+    x, sr = sf.read(test_file)
+    y = pv(x, alpha, win_type=win_type, win_size=win_size,
+           syn_hop_size=syn_hop_size, zero_pad=zero_pad,
+           restore_energy=restore_energy, fft_shift=fft_shift)
+
+    cmd = ['python', 'pytsmod/console.py', 'pv',
+           test_file, 'temp_cli.wav', str(alpha),
+           '-wt', win_type, '-ws', str(win_size),
+           '-sh', str(syn_hop_size), '-z', str(zero_pad),
+           '-e']
+    call(cmd)
+
+    sf.write('temp.wav', y, sr)
+    y_, _ = sf.read('temp.wav')
+
+    y_cli, _ = sf.read('temp_cli.wav')
+
+    os.remove('temp.wav')
+    os.remove('temp_cli.wav')
+
+    assert np.allclose(y_, y_cli)