Login | Register
Nerd ParadiseArtisanal tutorials since 1999
You may find the Python snippet at the end of this post useful for simple sound editing and effects. It contains a few helper functions and a class called Sound. This allows you to open, modify, and save WAVE files.

The Sound class has a field called "samples". This is a list containing 1 or 2 lists of numbers. These lists of numbers are the samples for each channel (1 channel is mono, 2 channels is stereo). By modifying the samples, you can edit the sound.

Reversing a stereo sound:
snd = create_sound_from_file('foo.wav')
snd.samples[0] = snd.samples[0][::-1# reverse left channel
snd.samples[1] = snd.samples[1][::-1# reverse right channel


Double the speed:
snd = create_sound_from_file('foo.wav')
snd.samples[0] = snd.samples[0][::2# cut out every other sample
snd.samples[1] = snd.samples[1][::2]


Save back out to file:
snd.save_to_file('foo2.wav')


And here's the magic code that makes it happen:
def push_int(byte_list, value, byte_length):
    while byte_length > 0:
        byte_list.append(chr(value & 255))
        value = value >> 8;
        byte_length -= 1
            
def push_string(byte_list, value):
    length = len(value)
    i = 0
    while i < length:
        byte_list.append(value[i])
        i += 1

class ByteStream:
    def __init__(self, file):
        self.i = 0
        c = open(file, 'rb')
        self.data = c.read()
        c.close()
        self.length = len(self.data)
        
    def has_next(self):
        return self.i < self.length
    
    def next_byte(self):
        output = ord(self.data[self.i])
        self.i += 1
        return output
    
    def next_string(self, length):
        output = ''
        while length > 0:
            output = self.data[self.i]
            self.i += 1
            length -= 1
        return output
    
    def next_int(self, byte_length):
        if byte_length == 1:
            output = ord(self.data[self.i])
            self.i += 1
        elif byte_length == 2:
            output = (ord(self.data[self.i]) +
                (ord(self.data[self.i + 1]) << 8))
            self.i += 2
        elif byte_length == 4:
            output = (ord(self.data[self.i]) +
                (ord(self.data[self.i + 1]) << 8) +
                (ord(self.data[self.i + 2]) << 16) +
                (ord(self.data[self.i + 3]) << 24))
            self.i += 4
        else:
            # how did this happen?
            pass
        return output

def create_sound_from_file(filename):
    bs = ByteStream(filename)
    
    ignore = bs.next_string(4# "RIFF"
    ignore = bs.next_int(4# file size
    ignore = bs.next_string(4# "WAVE"
    ignore = bs.next_string(4# "fmt "
    excess = bs.next_int(4) - 16 # file header size
    ignore = bs.next_int(2# 1 for PCM
    
    channels = bs.next_int(2# num channels
    
    sps = bs.next_int(4# samples per second (e.g. 44100)
    
    ignore = bs.next_int(4# bytes per second
    
    ignore = bs.next_int(2# bytes per sample
    
    bps = bs.next_int(2# bits per sample
    
    while excess > 0:
        ignore = bs.next_char()
        excess -= 1
    
    ignore = bs.next_string(4# "data"
    ignore = bs.next_int(4# length of data
    
    bytes_per_sample = bps // 8
    
    samples = []
    for channel in range(channels):
        samples.append([])
    
    while bs.has_next():
        channel_i = 0
        while channel_i < channels:
            samples[channel_i].append(bs.next_int(bytes_per_sample))
            channel_i += 1
    
    s = Sound(len(samples[0]), sps, channels, bps)
    s.samples = samples
    return s

class Sound:
    
    def __init__(self, sample_count, samples_per_second, channels, bits_per_sample):
        self.sps = samples_per_second
        self.bps = bits_per_sample
        self.samples = []
        for i in range(channels):
            self.samples.append([0] * sample_count)
    
    def save_to_file(self, filename):
        output = []
        
        push_string(output, "RIFF")
        
        header = self.generate_header()
        data = self.generate_data()
        
        push_int(output, len(data) + 364)
        
        output += header
        output += data
        
        c = open(filename, 'wb')
        c.write(''.join(output))
        c.close()

    def generate_header(self):
        output = []
        
        push_string(output, "WAVE")
        
        push_string(output, "fmt ")
        
        # Header size (16 bytes follow)
        push_int(output, 164)
        
        # 1 (to indicate PCM. Not sure what other valid values are)
        push_int(output, 12)
        
        # number of channels (stereo = 2, mono = 1)
        push_int(output, len(self.samples), 2)
        
        # samples per second
        push_int(output, self.sps, 4)
        
        # bytes per second (across all channels)
        push_int(output, self.bps / 8 * len(self.samples) * self.sps, 4)
        
        # bytes per sample across all channels
        push_int(output, self.bps / 8 * len(self.samples), 2)
        
        # bits per sample
        push_int(output, self.bps, 2)
        
        return output
    
    def generate_data(self):
        output = []
        
        samples = self.samples
        sample_i = 0
        samples_length = len(self.samples[0])
        num_channels = len(self.samples)
        bps_total = self.bps / 8
        while sample_i < samples_length:
            
            channels_i = 0
            while channels_i < num_channels:
                
                push_int(output, self.samples[channels_i][sample_i], bps_total)
                
                channels_i += 1
            
            sample_i += 1
        
        prefix = []
        push_string(prefix, 'data')
        
        # length of the rest of the data
        push_int(prefix, len(output), 4)
        
        return prefix + output