def main(argv):
import getopt
from wavestream import WaveReader
def usage():
print 'usage: %s [-M|-F] [-n pitchmin] [-m pitchmax] [-t threshold] wav ...' % argv[0]
return 100
try:
(opts, args) = getopt.getopt(argv[1:], 'MFn:m:t:')
except getopt.GetoptError:
return usage()
pitchmin = 70
pitchmax = 400
threshold = 0.9
for (k, v) in opts:
if k == '-M': (pitchmin,pitchmax) = (75,200) # male voice
elif k == '-F': (pitchmin,pitchmax) = (150,300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-t': threshold = float(v)
contour = None
for path in args:
src = WaveReader(path)
if contour is None:
contour = PitchContour(src.framerate,
pitchmin=pitchmin, pitchmax=pitchmax,
threshold=threshold)
contour.load(src.readraw(), src.nframes)
src.close()
for (t,p) in enumerate(contour.segments):
print t, p
return
def load_wav(self, wavpath, pitchpath):
src = WaveReader(wavpath)
fp = open(pitchpath)
for line in fp:
line = line.strip()
if not line:
print
(line, _, _) = line.partition('#')
if not line: continue
(f, _, pitch) = line.partition(' ')
f = int(f)
pitch = int(pitch)
src.seek(f)
r = []
(nframes, data) = src.readraw(int(src.framerate / pitch))
for (name, pat) in self.pats:
s = wavcorr.matchs16(pat, 0, nframes, data)
if self.threshold <= s:
r.append((s, name))
if r:
r.sort(reverse=True)
print f, ' '.join('%.04f:%s' % (s, name) for (s, name) in r)
fp.close()
src.close()
return
def fradosify(path, outfp, delta,
pitchmin=70, pitchmax=400, threshold=0.7):
print >>sys.stderr, 'reading: %r' % path
ratio = pow(2, delta/12.0)
src = WaveReader(path)
if src.nchannels != 1: raise ValueError('invalid number of channels')
if src.sampwidth != 2: raise ValueError('invalid sampling width')
contour = PitchContour(
src.framerate,
pitchmin=pitchmin, pitchmax=pitchmax,
threshold=threshold)
dst = WaveWriter(outfp, framerate=src.framerate)
nframes = src.nframes
buf = src.readraw(nframes)
contour.reset()
contour.load(buf, nframes)
def f(t):
x = contour.getavg(t)
if x != 0:
x = int(x*ratio)
return x
dst.writeraw(psola(buf, src.framerate, contour.getsrc, f, contour.wmin))
src.close()
dst.close()
return
def load_wav(self, wavpath, pitchpath):
src = WaveReader(wavpath)
fp = open(pitchpath)
for line in fp:
line = line.strip()
if not line:
print
(line,_,_) = line.partition('#')
if not line: continue
(f, _, pitch) = line.partition(' ')
f = int(f)
pitch = int(pitch)
src.seek(f)
r = []
(nframes, data) = src.readraw(int(src.framerate/pitch))
for (name,pat) in self.pats:
s = wavcorr.matchs16(pat, 0, nframes, data)
if self.threshold <= s:
r.append((s, name))
if r:
r.sort(reverse=True)
print f, ' '.join( '%.04f:%s' % (s,name) for (s,name) in r )
fp.close()
src.close()
return
def main(argv):
import getopt
from wavestream import WaveReader
def usage():
print 'usage: %s [-M|-F] [-n pitchmin] [-m pitchmax] [-t threshold] wav ...' % argv[
0]
return 100
try:
(opts, args) = getopt.getopt(argv[1:], 'MFn:m:t:')
except getopt.GetoptError:
return usage()
pitchmin = 70
pitchmax = 400
threshold = 0.9
for (k, v) in opts:
if k == '-M': (pitchmin, pitchmax) = (75, 200) # male voice
elif k == '-F': (pitchmin, pitchmax) = (150, 300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-t': threshold = float(v)
contour = None
for path in args:
src = WaveReader(path)
if contour is None:
contour = PitchContour(src.framerate,
pitchmin=pitchmin,
pitchmax=pitchmax,
threshold=threshold)
contour.load(src.readraw(), src.nframes)
src.close()
for (t, p) in enumerate(contour.segments):
print t, p
return
def fradosify(path, outfp, delta, pitchmin=70, pitchmax=400, threshold=0.7):
print >> sys.stderr, 'reading: %r' % path
ratio = pow(2, delta / 12.0)
src = WaveReader(path)
if src.nchannels != 1: raise ValueError('invalid number of channels')
if src.sampwidth != 2: raise ValueError('invalid sampling width')
contour = PitchContour(src.framerate,
pitchmin=pitchmin,
pitchmax=pitchmax,
threshold=threshold)
dst = WaveWriter(outfp, framerate=src.framerate)
nframes = src.nframes
buf = src.readraw(nframes)
contour.reset()
contour.load(buf, nframes)
def f(t):
x = contour.getavg(t)
if x != 0:
x = int(x * ratio)
return x
dst.writeraw(psola(buf, src.framerate, contour.getsrc, f, contour.wmin))
src.close()
dst.close()
return
def main(argv):
import getopt
def usage():
print 'usage: %s [-v] [-o out.wav] [-t f0-f1] wav ...' % argv[0]
return 100
def getv(v):
if '.' in v:
return float(v)
else:
return int(v)
try:
(opts, args) = getopt.getopt(argv[1:], 'vo:t:')
except getopt.GetoptError:
return usage()
verbose = 0
outfp = None
ranges = []
for (k, v) in opts:
if k == '-v': verbose += 1
elif k == '-o': outfp = open(v, 'wb')
elif k == '-t':
(f0,_,f1) = v.partition('-')
ranges.append((getv(f0), getv(f1)))
dst = None
for path in args:
src = WaveReader(path)
if verbose:
print >>sys.stderr, \
('%s: nchannels=%r, sampwidth=%r, framerate=%r, nframes=%r' %
(path, src.nchannels, src.sampwidth, src.framerate, src.nframes))
if dst is None:
if outfp is None:
dst = WavePlayer(nchannels=src.nchannels,
sampwidth=src.sampwidth,
framerate=src.framerate)
else:
dst = WaveWriter(outfp,
nchannels=src.nchannels,
sampwidth=src.sampwidth,
framerate=src.framerate)
for (f0,f1) in ranges:
if isinstance(f0, float):
f0 = int(f0*src.framerate)
if isinstance(f1, float):
f1 = int(f1*src.framerate)
src.seek(f0)
if f1 < f0:
f1 = src.nframes
(_,buf) = src.readraw(f1-f0)
dst.writeraw(buf)
src.close()
if dst is not None:
dst.close()
if outfp is not None:
outfp.close()
return
def main(argv):
import getopt
def usage():
print 'usage: %s [-M|-F] [-n pitchmin] [-m pitchmax] [-t threshold] wav ...' % argv[
0]
return 100
try:
(opts, args) = getopt.getopt(argv[1:], 'MFn:m:t:')
except getopt.GetoptError:
return usage()
pitchmin = 70
pitchmax = 400
threshold = 0.9
bufsize = 10000
for (k, v) in opts:
if k == '-M': (pitchmin, pitchmax) = (75, 200) # male voice
elif k == '-F': (pitchmin, pitchmax) = (150, 300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-t': threshold = float(v)
detector = None
for path in args:
src = WaveReader(path)
if detector is None:
detector = PitchDetector(src.framerate,
pitchmin=pitchmin,
pitchmax=pitchmax)
i = 0
while 1:
(nframes, buf) = src.readraw(bufsize)
if not nframes: break
pitches = detector.feed(buf, nframes)
for (n, t, freq, mag, data) in pitches:
if threshold <= t:
print i, n, t, freq, mag
else:
print i, n, t
i += n
src.close()
return
def fradosify(path, outfp, delta):
print >>sys.stderr, 'reading: %r' % path
ratio = pow(2, delta/12.0)
src = WaveReader(path)
if src.nchannels != 1: raise ValueError('invalid number of channels')
if src.sampwidth != 2: raise ValueError('invalid sampling width')
dst = WaveWriter(outfp, framerate=src.framerate)
contour = PitchContour(src.framerate)
contour.load(src)
def f(t):
x = contour.getavg(t)
if x is not None:
x = int(x*ratio)
return x
src.seek(0)
dst.write(psola(src.read(), src.framerate, contour.getsrc, f, contour.wmin))
src.close()
dst.close()
return
def main(argv):
import getopt
def usage():
print 'usage: %s [-M|-F] [-n pitchmin] [-m pitchmax] [-t threshold] wav ...' % argv[0]
return 100
try:
(opts, args) = getopt.getopt(argv[1:], 'MFn:m:t:')
except getopt.GetoptError:
return usage()
pitchmin = 70
pitchmax = 400
threshold = 0.9
bufsize = 10000
for (k, v) in opts:
if k == '-M': (pitchmin,pitchmax) = (75,200) # male voice
elif k == '-F': (pitchmin,pitchmax) = (150,300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-t': threshold = float(v)
detector = None
for path in args:
src = WaveReader(path)
if detector is None:
detector = PitchDetector(src.framerate,
pitchmin=pitchmin, pitchmax=pitchmax)
i = 0
while 1:
(nframes,buf) = src.readraw(bufsize)
if not nframes: break
pitches = detector.feed(buf, nframes)
for (n,t,freq,mag,data) in pitches:
if threshold <= t:
print i,n,t,freq,mag
else:
print i,n,t
i += n
src.close()
return
def load_pat(self, path, name=None):
fp = WaveReader(path)
(_, pat) = fp.readraw()
self.pats.append((name or path, pat))
fp.close()
return
def main(argv):
import getopt
import fileinput
def usage():
print 'usage: %s [-v] [-o out.wav] [script ...]' % argv[0]
return 100
def getv(v):
try:
if '.' in v:
return float(v)
else:
return int(v)
except ValueError:
return 0
try:
(opts, args) = getopt.getopt(argv[1:], 'vo:')
except getopt.GetoptError:
return usage()
verbose = 0
outfp = None
for (k, v) in opts:
if k == '-v': verbose += 1
elif k == '-o': outfp = open(v, 'wb')
#
if not args: return usage()
path = args.pop(0)
src = WaveReader(path)
#
waves = []
for line in fileinput.input(args):
(line,_,_) = line.partition('#')
line = line.strip()
if not line: continue
(t,dur,path) = line.split('\t')
t = getv(t)
dur = getv(dur)
wav = WaveReader(path)
assert wav.nchannels == src.nchannels
assert wav.sampwidth == src.sampwidth
#assert wav.framerate == src.framerate
if isinstance(t, float):
t = int(t*src.framerate)
if isinstance(dur, float):
dur = int(dur*src.framerate)
buf = wav.read(dur)
wav.close()
waves.append((t, buf))
waves.append((src.nframes, []))
#
if outfp is not None:
dst = WaveWriter(outfp,
nchannels=src.nchannels,
sampwidth=src.sampwidth,
framerate=src.framerate)
else:
dst = WavePlayer(nchannels=src.nchannels,
sampwidth=src.sampwidth,
framerate=src.framerate)
#
t0 = 0
bufs = []
for (t1,buf1) in sorted(waves, key=lambda (t,_): t):
dt = (t1-t0)*dst.nchannels
tmp = [src.read(t1-t0)]
assert len(tmp[0]) == dt
for (i,b) in enumerate(bufs):
if dt <= len(b):
tmp.append(b[:dt])
else:
tmp.append(b+[0]*(dt-len(b)))
bufs[i] = b[dt:]
bufs.append(buf1)
bufs = [ b for b in bufs if b ]
dst.write(mix(tmp))
t0 = t1
#
dst.close()
if outfp is not None:
outfp.close()
return
def main(argv):
import getopt
from wavestream import WaveReader
def usage():
print(
'usage: %s [-d] [-M|-F] [-n pitchmin] [-m pitchmax]'
' [-T threshold_sim] [-S threshold_mag] wav ...' % argv[0])
return 100
def parse_range(x):
(b, _, e) = x.partition('-')
try:
b = int(b)
except ValueError:
b = 0
try:
e = int(e)
except ValueError:
e = 0
return (b, e)
try:
(opts, args) = getopt.getopt(argv[1:], 'dMFn:m:T:S:')
except getopt.GetoptError:
return usage()
debug = 0
pitchmin = 70
pitchmax = 400
threshold_sim = 0.75
threshold_mag = 0.025
bufsize = 10000
for (k, v) in opts:
if k == '-d': debug += 1
elif k == '-M': (pitchmin, pitchmax) = (75, 200) # male voice
elif k == '-F': (pitchmin, pitchmax) = (150, 300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-T': threshold_sim = float(v)
elif k == '-S': threshold_mag = float(v)
detector = None
smoother = None
for arg1 in args:
(path, _, ranges) = arg1.partition(':')
ranges = [parse_range(x) for x in ranges.split(',') if x]
if not ranges:
ranges.append((0, 0))
print '#', path, ranges
src = WaveReader(path)
if src.nchannels != 1: raise ValueError('invalid number of channels')
if src.sampwidth != 2: raise ValueError('invalid sampling width')
framerate = src.framerate
if detector is None:
detector = PitchDetector(wmin=framerate / pitchmax,
wmax=framerate / pitchmin,
threshold_sim=threshold_sim)
smoother = PitchSmoother(2 * framerate / pitchmin,
threshold_sim=threshold_sim,
threshold_mag=threshold_mag)
for (b, e) in ranges:
if e == 0:
e = src.nframes
src.seek(b)
length = e - b
i0 = b
while length:
(nframes, buf) = src.readraw(min(bufsize, length))
if not nframes: break
length -= nframes
seq = detector.feed(buf, nframes)
for (n0, pitches, data) in seq:
if debug:
print('# %d: %r' % (n0, pitches))
for streak in smoother.feed(i0, n0, pitches):
for (i1, n1, spitches) in streak:
print i1, n1, ' '.join('%d:%.4f' % (w, sim)
for (_, w, sim) in spitches)
print
i0 += n0
src.close()
return
def main(argv):
import getopt
from wavestream import WaveReader
def usage():
print ('usage: %s [-M|-F] [-n pitchmin] [-m pitchmax] [-t threshold] '
'[-o out.mid] [-w wsize] [-p instru] wav ...' % argv[0])
return 100
try:
(opts, args) = getopt.getopt(argv[1:], 'MFn:m:t:o:w:p:')
except getopt.GetoptError:
return usage()
pitchmin = 70
pitchmax = 400
threshold = 0.95
outpath = 'out.mid'
wsize = 50
instru = 0
attack = 70
release = 70
for (k, v) in opts:
if k == '-M': (pitchmin,pitchmax) = (75,200) # male voice
elif k == '-F': (pitchmin,pitchmax) = (150,300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-t': threshold = float(v)
elif k == '-o': outpath = v
elif k == '-w': wsize = int(v)
elif k == '-p': instru = int(v)
contour = None
for path in args:
src = WaveReader(path)
if contour is None:
contour = PitchContour(src.framerate,
pitchmin=pitchmin, pitchmax=pitchmax,
threshold=threshold)
contour.load(src.readraw(), src.nframes)
src.close()
#
events = [midi.ProgramChangeEvent(tick=0, channel=0, data=[instru])]
window = []
km0 = 0
kt = 0
kd = 0
for p in contour.segments:
if p == 0:
k = 0
else:
i = getpt(FRANGES, p)
(_,k) = FRANGES[i-1]
#print '%d/%d' % (p,k),
window.append(k)
if len(window) < wsize: continue
window = window[1:]
km1 = majority(window)
if km0 == km1:
kd += 1
else:
print km0, kd
if km0 == 0:
kt += kd
else:
events.append(midi.NoteOnEvent(tick=kt, channel=0, data=[km0, attack]))
events.append(midi.NoteOffEvent(tick=kd, channel=0, data=[km0, release]))
kt = 0
kd = 0
km0 = km1
events.append(midi.EndOfTrackEvent(tick=0, data=[]))
pat = midi.Pattern(tracks=[events])
midi.write_midifile(outpath, pat)
return
def main(argv):
import getopt
import fileinput
def usage():
print 'usage: %s [-v] [-o out.wav] [script ...]' % argv[0]
return 100
def getv(v):
try:
if '.' in v:
return float(v)
else:
return int(v)
except ValueError:
return 0
try:
(opts, args) = getopt.getopt(argv[1:], 'vo:')
except getopt.GetoptError:
return usage()
verbose = 0
outfp = None
for (k, v) in opts:
if k == '-v': verbose += 1
elif k == '-o': outfp = open(v, 'wb')
#
if not args: return usage()
path = args.pop(0)
src = WaveReader(path)
#
waves = []
for line in fileinput.input(args):
(line, _, _) = line.partition('#')
line = line.strip()
if not line: continue
(t, dur, path) = line.split('\t')
t = getv(t)
dur = getv(dur)
wav = WaveReader(path)
assert wav.nchannels == src.nchannels
assert wav.sampwidth == src.sampwidth
#assert wav.framerate == src.framerate
if isinstance(t, float):
t = int(t * src.framerate)
if isinstance(dur, float):
dur = int(dur * src.framerate)
buf = wav.read(dur)
wav.close()
waves.append((t, buf))
waves.append((src.nframes, []))
#
if outfp is not None:
dst = WaveWriter(outfp,
nchannels=src.nchannels,
sampwidth=src.sampwidth,
framerate=src.framerate)
else:
dst = WavePlayer(nchannels=src.nchannels,
sampwidth=src.sampwidth,
framerate=src.framerate)
#
t0 = 0
bufs = []
for (t1, buf1) in sorted(waves, key=lambda (t, _): t):
dt = (t1 - t0) * dst.nchannels
tmp = [src.read(t1 - t0)]
assert len(tmp[0]) == dt
for (i, b) in enumerate(bufs):
if dt <= len(b):
tmp.append(b[:dt])
else:
tmp.append(b + [0] * (dt - len(b)))
bufs[i] = b[dt:]
bufs.append(buf1)
bufs = [b for b in bufs if b]
dst.write(mix(tmp))
t0 = t1
#
dst.close()
if outfp is not None:
outfp.close()
return
def main(argv):
import getopt
from wavestream import WaveReader
def usage():
print(
'usage: %s [-M|-F] [-n pitchmin] [-m pitchmax] [-t threshold] '
'[-o out.mid] [-w wsize] [-p instru] wav ...' % argv[0])
return 100
try:
(opts, args) = getopt.getopt(argv[1:], 'MFn:m:t:o:w:p:')
except getopt.GetoptError:
return usage()
pitchmin = 70
pitchmax = 400
threshold = 0.97
outpath = 'out.mid'
wsize = 50
instru = 0
attack = 70
release = 70
for (k, v) in opts:
if k == '-M': (pitchmin, pitchmax) = (75, 200) # male voice
elif k == '-F': (pitchmin, pitchmax) = (150, 300) # female voice
elif k == '-n': pitchmin = int(v)
elif k == '-m': pitchmax = int(v)
elif k == '-t': threshold = float(v)
elif k == '-o': outpath = v
elif k == '-w': wsize = int(v)
elif k == '-p': instru = int(v)
contour = None
for path in args:
src = WaveReader(path)
if contour is None:
contour = PitchContour(src.framerate,
pitchmin=pitchmin,
pitchmax=pitchmax,
threshold=threshold)
contour.load(src.readraw(), src.nframes)
src.close()
events = [midi.ProgramChangeEvent(tick=0, channel=0, data=[instru])]
window = []
km0 = 0
kt = 0
kd = 0
piano = open("../server/piano.txt", "w")
for p in contour.segments:
if p == 0:
k = 0
else:
i = getpt(FRANGES, p)
(_, k) = FRANGES[i - 1]
window.append(k)
if len(window) < wsize: continue
window = window[1:]
km1 = majority(window)
if km0 == km1:
kd += 1
else:
print km0, kd
piano.write(str(km0) + ' ' + str(kd) + '\r\n')
if km0 == 0:
kt += kd
else:
events.append(
midi.NoteOnEvent(tick=kt, channel=0, data=[km0, attack]))
events.append(
midi.NoteOffEvent(tick=kd, channel=0, data=[km0, release]))
kt = 0
kd = 0
km0 = km1
piano.write('EOF')
piano.close()
events.append(midi.EndOfTrackEvent(tick=0, data=[]))
pat = midi.Pattern(tracks=[events])
midi.write_midifile(outpath, pat)
return
def load_pat(self, path, name=None):
fp = WaveReader(path)
(_, pat) = fp.readraw()
self.pats.append((name or path, pat))
fp.close()
return