pratool / phase-vocoder Goto Github PK

import pyaudio
import wave
import time
import numpy as np
import struct


def speedx(snd_array, factor):
    """ Speeds up / slows down a sound, by some factor. """
    indices = np.round(np.arange(0, len(snd_array), factor))
    indices = indices[indices < len(snd_array)].astype(int)
    return snd_array[indices]


def stretch(snd_array, factor, window_size, h):
    """ Stretches/shortens a sound, by some factor. """
    phase = np.zeros(window_size)
    hanning_window = np.hanning(window_size)
    result = np.zeros(len(snd_array) / factor + window_size)

    for i in np.arange(0, len(snd_array) - (window_size + h), h*factor):
        # Two potentially overlapping subarrays
        a1 = snd_array[i: i + window_size]
        a2 = snd_array[i + h: i + window_size + h]

        # The spectra of these arrays
        s1 = np.fft.fft(hanning_window * a1)
        s2 = np.fft.fft(hanning_window * a2)

        # Rephase all frequencies
        phase = (phase + np.angle(s2/s1)) % 2*np.pi

        a2_rephased = np.fft.ifft(np.abs(s2)*np.exp(1j*phase))
        i2 = int(i/factor)
        result[i2: i2 + window_size] += hanning_window*a2_rephased.real

    # normalize (16bit)
    result = ((2**(16-4)) * result/result.max())

    return result.astype('int16')


def pitchshift(snd_array, n, window_size=2**13, h=2**11):
    """ Changes the pitch of a sound by ``n`` semitones. """
    factor = 2**(2.0 * n / 12.0)
    stretched = stretch(snd_array, 1.0/factor, window_size, h)
    return speedx(stretched[window_size:], factor)




wf = wave.open("/Users/miooo/Desktop/CHP_Glissandos_MinorSamples/Min_dn_sl_F_f.wav", 'rb')

# instantiate PyAudio (1)
p = pyaudio.PyAudio()

# open stream (2)
stream = p.open(format=p.get_format_from_width(wf.getsampwidth()),
                channels=wf.getnchannels(),
                rate=wf.getframerate(),
                output=True)

# read data
data = wf.readframes(1024)
data = np.fromstring(data, np.int16)

# play stream (3)
while len(data) > 0:
    new_data = pitchshift(data, 1)
    print new_data
    out_data = new_data.astype(np.int16).tostring()
    print out_data
    stream.write(out_data)
    data = wf.readframes(1024)
    data = np.fromstring(data, np.int16)

# stop stream (4)
stream.stop_stream()
stream.close()

# close PyAudio (5)
p.terminate()