if __name__ == '__main__':
global dt
# figure out the correct dt length based on the tempo of the output wave you want
dt = compute_dt(bpm)
# read and parse each sample wave
for i in range(1, len(sys.argv) - 2):
#read in the wave file and unpack its data
wave_data = waveIO.read_wav_file(sys.argv[i])
wave_data = waveIO.unpack(wave_data)
wave_time = len(wave_data) / sr
note_length = dt * sr
wave_chunks = split_wave(wave_data, int(note_length))
#test_sample_wave(wave_chunks)
for i in range(len(wave_chunks)):
chunk = wave_chunks[i]
if len(chunk) != int(note_length):
continue
store_note(chunk)
print_notes()
plot_notes()
# read and parse the music file
musicfile = sys.argv[-2]
song = build_song(musicfile, dt)
waveIO.write_wav_file(sys.argv[-1], waveIO.pack(song))
wave_chunks = [] #Instead of multiplying the entire wave by the windowing function, we can simply examine window-sized pieces of the wave at a time #counting by WINDOW_LENGTH*(1-OVERLAP_PERCENT) allows adjustment of what percent of each window overlaps with the next #Example: OVERLAP_PERCENT = 0 means the next window will start where the previous window ended #Example: OVERLAP_PERCENT = 0.5 means the next window will start at the halfway mark of the previous window #This is necessary when using nonrectangular window functions, to prevent loss of data for i in range(0, len(wave_data), int(WINDOW_LENGTH*(1-OVERLAP_PERCENT))): window = [] window.extend(wave_data[i:i+WINDOW_LENGTH]) #multiply by the window function window = WINDOW_FUNC(window) wave_chunks.append(window) for i in range(len(wave_chunks)): chunk = wave_chunks[i] if len(chunk) != WINDOW_LENGTH: continue store_note(chunk) print_notes() plot_notes() # read and parse the music file musicfile = sys.argv[-2] song = build_song(musicfile, WINDOW_LENGTH / float(sr)) waveIO.write_wav_file(sys.argv[-1], waveIO.pack(song))
"A":440.,
"A#":466.16,
"B":493.88,
"C":261.63,
"C#":277.18,
"D":293.66,
"D#":311.13,
"E":329.63,
"F":349.23,
"F#":369.99,
"G":391.99,
"G#":415.31
}
freqs5 = {}
freqs6 = {}
note_length = 8938
if __name__ == '__main__':
for note in freqs4:
freqs5[note] = 2*freqs4[note]
freqs6[note] = 4*freqs4[note]
musicfile = sys.argv[1]
musicwave = build_song(musicfile)
data, freqs, bins, im = plt.specgram(musicwave, NFFT=2048, Fs=44100, noverlap=900)
plt.ylim(0,800)
plt.title("Spectrogram of analytic_furelise_hanning.wav")
plt.show()
waveIO.write_wav_file("analytic_furelise.wav", waveIO.pack(musicwave))
"C": 261.63,
"C#": 277.18,
"D": 293.66,
"D#": 311.13,
"E": 329.63,
"F": 349.23,
"F#": 369.99,
"G": 391.99,
"G#": 415.31
}
freqs5 = {}
freqs6 = {}
note_length = 8938
if __name__ == '__main__':
for note in freqs4:
freqs5[note] = 2 * freqs4[note]
freqs6[note] = 4 * freqs4[note]
musicfile = sys.argv[1]
musicwave = build_song(musicfile)
data, freqs, bins, im = plt.specgram(musicwave,
NFFT=2048,
Fs=44100,
noverlap=900)
plt.ylim(0, 800)
plt.title("Spectrogram of analytic_furelise_hanning.wav")
plt.show()
waveIO.write_wav_file("analytic_furelise.wav", waveIO.pack(musicwave))
plt.xlim(0,1000)
plt.xlabel("Frequency (Hz)")
plt.ylabel("Absolute value of FFT Magnitude, divided by N")
#plt.yscale("log")
plt.title("Power spectrum of sample wave")
plt.show()
data, freqs, bins, im = plt.specgram(t_sin, NFFT=NFFT, Fs=fs, noverlap=NFFT/2)
plt.title("Spectrogram, notes in reverse, window size = 1024, sr = 44100")
plt.ylabel("Frequency (Hz)")
plt.xlabel("Time (s)")
plt.ylim(0,800)
plt.show()
#print(data)
waveIO.write_wav_file("samplewave.wav", waveIO.pack(t_sin))
##############
# Now build a universal sample, with short sections for each note through a few octaves
t = numpy.arange(0, 0.5, dt)
big_sample = create_wave(440, t)
big_sample = numpy.append(big_sample, create_wave(261, t))
big_sample = numpy.append(big_sample, create_wave(277, t))
big_sample = numpy.append(big_sample, create_wave(293, t))
big_sample = numpy.append(big_sample, create_wave(311, t))
big_sample = numpy.append(big_sample, create_wave(329, t))
big_sample = numpy.append(big_sample, create_wave(349, t))
big_sample = numpy.append(big_sample, create_wave(369, t))
big_sample = numpy.append(big_sample, create_wave(391, t))
big_sample = numpy.append(big_sample, create_wave(415, t))
plt.ylabel("Absolute value of FFT Magnitude, divided by N")
#plt.yscale("log")
plt.title("Power spectrum of sample wave")
plt.show()
data, freqs, bins, im = plt.specgram(t_sin,
NFFT=NFFT,
Fs=fs,
noverlap=NFFT / 2)
plt.title("Spectrogram, notes in reverse, window size = 1024, sr = 44100")
plt.ylabel("Frequency (Hz)")
plt.xlabel("Time (s)")
plt.ylim(0, 800)
plt.show()
#print(data)
waveIO.write_wav_file("samplewave.wav", waveIO.pack(t_sin))
##############
# Now build a universal sample, with short sections for each note through a few octaves
t = numpy.arange(0, 0.5, dt)
big_sample = create_wave(440, t)
big_sample = numpy.append(big_sample, create_wave(261, t))
big_sample = numpy.append(big_sample, create_wave(277, t))
big_sample = numpy.append(big_sample, create_wave(293, t))
big_sample = numpy.append(big_sample, create_wave(311, t))
big_sample = numpy.append(big_sample, create_wave(329, t))
big_sample = numpy.append(big_sample, create_wave(349, t))
big_sample = numpy.append(big_sample, create_wave(369, t))
big_sample = numpy.append(big_sample, create_wave(391, t))
big_sample = numpy.append(big_sample, create_wave(415, t))
big_sample = numpy.append(big_sample, create_wave(466.16, t))
