Small fixes

- Added some info about Windows support to README
- Fixed plots in demo.py
- Changed demo.py to use PySoundFile (libsndfile) instead of scipy for
reading/writing .wav files

Author: wuaalb

.gitignore

@@ -4,3 +4,5 @@ World/
 src/
 *.cpp
 *.so
+*.pyd
+*.egg-info

README.md

@@ -1,4 +1,4 @@
-# PyWorldVocoder<br/>A Python wrapper for World Vocoder
+# PyWorldVocoder - A Python wrapper for World Vocoder
 
 
 | **`Linux`** |
@@ -9,17 +9,15 @@
 Morise's World Vocoder is a fast and high-quality vocoder.
 World Vocoder parameterizes speech into three components:
 
-  5. Pitch (fundamental frequency, F0) contour  
-  2. smoothed spectrogram  
-  3. aperiodicity  
+  1. Pitch (fundamental frequency, F0) contour  
+  2. Harmonic spectral envelope
+  3. Aperiodic spectral envelope (relative to the harmonic spectral envelope)
 
 It can also resynthesize speech using these features (see examples below).
 
 For more information, please visit [Morise's World repository](https://github.com/mmorise/World) 
 and the [official website of World Vocoder](http://ml.cs.yamanashi.ac.jp/world/english/)
 
-<br/>
-
 
 ## I. APIs
 
@@ -41,20 +39,20 @@ y = pw.synthesize(f0, sp, ap, fs, pyDioOpt.option['frame_period'])
 f0, sp, ap, pyDioOpt = pw.wav2world(x, fs)
 ```
 
-<br/>
 
 ## II. Installation
 ### Environment/Dependencies  
-- Linux Ubuntu 14.04/16.04  
-- Python 2.7.6/3.5  
-- Cython 0.24 (or later versions; required)  
-- Scipy  
-- Numpy  
-- argparse  
-- Matplotlib (optional; for demo.py only)  
+- Linux Ubuntu 14.04/16.04, Windows
+- Python 2.7/3.5
+- Cython 0.24 (or later versions; required)
+- Numpy
+- argparse (optional; for demo.py only)
+- PySoundFile (optional; for demo.py only)
+- Matplotlib (optional; for demo.py only)
 
 You can simply install these by `pip install -r requirements.txt`
 
+
 ### Installation procedures
 ```bash
 pip install -U pip
@@ -64,49 +62,48 @@ cd Python-Wrapper-for-World-Vocoder
 bash download_vocoder.sh
 python setup.py install
 ```
-It will automatically `git clone` Morise's World Vocoder (C++ version).  
-<br/>
+It will automatically `git clone` Morise's World Vocoder (C++ version).<br/>
+Alternatively you can clone or download the World repository manually and copy its "src" directory to this repositories directory.<br/>
 As for installation mode (the last line), you can choose from the following options.
 
 
 ### Installation Mode
-9. If you want to "install" this package, try <br/>
-  `python setup.py install`  
-  (add `--user` if you don't have root access)
-0. If you just want to try out some experiments, execute  
-  `python setup.py build_ext --inplace` <br/>
+1. If you want to "install" this package, try<br/>
+   `python setup.py install`<br/>
+   (add `--user` if you don't have root access)
+2. If you just want to try out some experiments, execute<br/>
+  `python setup.py build_ext --inplace`<br/>
   Then you can use PyWorld from this directory.<br/>
-  You can also copy the resulting **pyworld.so** file to  
-  `~/.local/lib/python2.7/site-packages` <br/>
-  so that you can use it everywhere like an installed package.
+  You can also copy the resulting **pyworld.so** (pyworld.{arch}.pyd on Windows) file to
+  `~/.local/lib/python2.7/site-packages` (or corresponding Windows directory)
+  so that you can use it everywhere like an installed package.<br/>
+  Alternatively you can copy/symlink the compiled files using pip, e.g. `pip install -e .`
 
 
 ### Validation
-You can validate installation by running     
-`python demo.py`  
+You can validate installation by running
+`python demo.py`
 to see if you get results in `test/` direcotry.
 
-<br/>
 
 ## Troubleshooting
-0. Upgrade your Cython version to 0.24.<br/>
+1. Upgrade your Cython version to 0.24.<br/>
    (I failed to build it on Cython 0.20.1post0)<br/>
-   It'll require you to download Cython form http://cython.org/ <br/>
+   It'll require you to download Cython form http://cython.org/<br/>
    Unzip it, and `python setup.py install` it.<br/>
    (I tried `pip install Cython` but the upgrade didn't seem correct)<br/>
-   (Again, add `--user` if you don't have root access.)  
+   (Again, add `--user` if you don't have root access.)
+2. The following code might be needed in some configurations:
 
-1. The following code might be needed in some configurations:
-  ```python
-  import matplotlib  
-  matplotlib.use('Agg')
-  ```
+ ```python
+ import matplotlib
+ matplotlib.use('Agg')
+ ```
 
-<br/>
 
 ## Note:
 1. This wrapper is an updated version of sotelo's "world.py"<br/>
-  https://github.com/sotelo/world.py
+   https://github.com/sotelo/world.py
 
 ## TODO List
 

demo.py

@@ -1,39 +1,27 @@
-from __future__ import print_function
-import pyworld as pw
+from __future__ import division, print_function
+
 import os
 from shutil import rmtree
+import argparse
+
 import numpy as np
-import matplotlib
-matplotlib.use('Agg')
+#import matplotlib
+#matplotlib.use('Agg')
 import matplotlib.pyplot as plt
-from scipy.io.wavfile import read, write
-import argparse
+
+import soundfile as sf
+import pyworld as pw
+
 
 parser = argparse.ArgumentParser()
 parser.add_argument("-r", "--frame_rate", type=int, default=5)
 parser.add_argument("-s", "--speed", type=int, default=1)
 
-# =================================
-SHORT_MAX = 32767
-EPSILON = 1e-8
-def wavread(filename):
-    fs, x = read(filename)
-    x = x.astype(np.float) / SHORT_MAX
-    return x, fs
-
-
-def wavwrite(filename, fs, y):
-    ymax = np.max(np.abs(y))
-    if ymax < 1.0:
-        y = y * SHORT_MAX
-    else:
-        y = (y / ymax) * SHORT_MAX
-    y = y.astype(np.int16)
-    write(filename, fs, y)
 
+EPSILON = 1e-8
 
 def savefig(filename, figlist, log=True):
-    h = 10
+    #h = 10
     n = len(figlist)
     # peek into instances
     f = figlist[0]
@@ -46,26 +34,27 @@ def savefig(filename, figlist, log=True):
                 plt.xlim([0, len(f)])
     elif len(f.shape) == 2:
         Nsmp, dim = figlist[0].shape
-        figsize=(h * float(Nsmp) / dim, len(figlist) * h)
-        plt.figure(figsize=figsize)
+        #figsize=(h * float(Nsmp) / dim, len(figlist) * h)
+        #plt.figure(figsize=figsize)
+        plt.figure()
         for i, f in enumerate(figlist):
             plt.subplot(n, 1, i+1)
             if log:
-                plt.imshow(np.log(f.T + EPSILON))
+                x = np.log(f + EPSILON)
             else:
-                plt.imshow(f.T + EPSILON)
+                x = f + EPSILON
+            plt.imshow(x.T, origin='lower', interpolation='none', aspect='auto', extent=(0, x.shape[0], 0, x.shape[1]))
     else:
         raise ValueError('Input dimension must < 3.')
     plt.savefig(filename)
-# =================================
 
 
 def main(args):
     if os.path.isdir('test'):
         rmtree('test')
     os.mkdir('test')
 
-    x, fs = wavread('utterance/vaiueo2d.wav')
+    x, fs = sf.read('utterance/vaiueo2d.wav')
 
     # 1. A convient way
     f0, sp, ap, pyDioOpt = pw.wav2world(x, fs)    # use default options
@@ -84,14 +73,14 @@ def main(args):
     _sp = pw.cheaptrick(x, _f0, t, fs)
     _ap = pw.d4c(x, _f0, t, fs)
     _y = pw.synthesize(_f0, _sp, _ap, fs, pyDioOpt.option['frame_period'])
-    wavwrite('test/y_without_f0_refinement.wav', fs, _y)
+    sf.write('test/y_without_f0_refinement.wav', _y, fs)
 
     # 2-2 With F0 refinement (using stonemask)
     f0 = pw.stonemask(x, _f0, t, fs)
     sp = pw.cheaptrick(x, f0, t, fs)
     ap = pw.d4c(x, f0, t, fs)
     y = pw.synthesize(f0, sp, ap, fs, pyDioOpt.option['frame_period'])
-    wavwrite('test/y_with_f0_refinement.wav', fs, y)
+    sf.write('test/y_with_f0_refinement.wav', y, fs)
 
     # Comparison
     savefig('test/wavform.png', [x, _y, y])

requirements.txt

@@ -1,5 +1,5 @@
 numpy
 matplotlib
-scipy
 argparse
 cython
+pysoundfile