python/scripts/aubiocut

   1 #! /usr/bin/env python
   2
   3 """ this file was written by Paul Brossier
   4   it is released under the GNU/GPL license.
   5 """
   6
   7 import sys
   8 #from aubio.task import *
   9
  10 usage = "usage: %s [options] -i soundfile" % sys.argv[0]
  11 usage += "\n help: %s -h" % sys.argv[0]
  12
  13 def parse_args():
  14     from optparse import OptionParser
  15     parser = OptionParser(usage=usage)
  16     parser.add_option("-i", "--input", action = "store", dest = "source_file",
  17             help="input sound file to analyse", metavar = "<source_file>")
  18     parser.add_option("-m","--method",
  19             action="store", dest="onset_method", default='default',
  20             metavar = "<onset_method>",
  21             help="onset detection method [default=default] \
  22                     complexdomain|hfc|phase|specdiff|energy|kl|mkl")
  23     # cutting methods
  24     parser.add_option("-b","--beat",
  25             action="store_true", dest="beat", default=False,
  26             help="use beat locations")
  27     """
  28     parser.add_option("-S","--silencecut",
  29             action="store_true", dest="silencecut", default=False,
  30             help="use silence locations")
  31     parser.add_option("-s","--silence",
  32             metavar = "<value>",
  33             action="store", dest="silence", default=-70,
  34             help="silence threshold [default=-70]")
  35             """
  36     # algorithm parameters
  37     parser.add_option("--samplerate",
  38             metavar = "<freq>", type='int',
  39             action="store", dest="samplerate", default=0,
  40             help="samplerate at which the file should be represented")
  41     parser.add_option("-B","--bufsize",
  42             action="store", dest="bufsize", default=512,
  43             metavar = "<size>", type='int',
  44             help="buffer size [default=512]")
  45     parser.add_option("-H","--hopsize",
  46             metavar = "<size>", type='int',
  47             action="store", dest="hopsize", default=256,
  48             help="overlap size [default=256]")
  49     parser.add_option("-t","--threshold",
  50             metavar = "<value>", type="float",
  51             action="store", dest="threshold", default=0.3,
  52             help="onset peak picking threshold [default=0.3]")
  53     parser.add_option("-c","--cut",
  54             action="store_true", dest="cut", default=False,
  55             help="cut input sound file at detected labels \
  56                     best used with option -L")
  57     """
  58     parser.add_option("-D","--delay",
  59             action = "store", dest = "delay", type = "float",
  60             metavar = "<seconds>", default=0,
  61             help="number of seconds to take back [default=system]\
  62                     default system delay is 3*hopsize/samplerate")
  63     parser.add_option("-C","--dcthreshold",
  64             metavar = "<value>",
  65             action="store", dest="dcthreshold", default=1.,
  66             help="onset peak picking DC component [default=1.]")
  67     parser.add_option("-M","--mintol",
  68             metavar = "<value>",
  69             action="store", dest="mintol", default=0.048,
  70             help="minimum inter onset interval [default=0.048]")
  71     parser.add_option("-L","--localmin",
  72             action="store_true", dest="localmin", default=False,
  73             help="use local minima after peak detection")
  74     parser.add_option("-d","--derivate",
  75             action="store_true", dest="derivate", default=False,
  76             help="derivate onset detection function")
  77     parser.add_option("-z","--zerocross",
  78             metavar = "<value>",
  79             action="store", dest="zerothres", default=0.008,
  80             help="zero-crossing threshold for slicing [default=0.00008]")
  81             """
  82     # plotting functions
  83     """
  84     parser.add_option("-p","--plot",
  85             action="store_true", dest="plot", default=False,
  86             help="draw plot")
  87     parser.add_option("-x","--xsize",
  88             metavar = "<size>",
  89             action="store", dest="xsize", default=1.,
  90             type='float', help="define xsize for plot")
  91     parser.add_option("-y","--ysize",
  92             metavar = "<size>",
  93             action="store", dest="ysize", default=1.,
  94             type='float', help="define ysize for plot")
  95     parser.add_option("-f","--function",
  96             action="store_true", dest="func", default=False,
  97             help="print detection function")
  98     parser.add_option("-n","--no-onsets",
  99             action="store_true", dest="nplot", default=False,
 100             help="do not plot detected onsets")
 101     parser.add_option("-O","--outplot",
 102             metavar = "<output_image>",
 103             action="store", dest="outplot", default=None,
 104             help="save plot to output.{ps,png}")
 105     parser.add_option("-F","--spectrogram",
 106             action="store_true", dest="spectro", default=False,
 107             help="add spectrogram to the plot")
 108     """
 109     parser.add_option("-v","--verbose",
 110             action="store_true", dest="verbose", default=True,
 111             help="make lots of noise [default]")
 112     parser.add_option("-q","--quiet",
 113             action="store_false", dest="verbose", default=True,
 114             help="be quiet")
 115     (options, args) = parser.parse_args()
 116     if not options.source_file:
 117         print "no file name given\n", usage
 118         sys.exit(1)
 119     return options, args
 120
 121 if __name__ == '__main__':
 122     options, args = parse_args()
 123
 124     hopsize = options.hopsize
 125     bufsize = options.bufsize
 126     samplerate = options.samplerate
 127     source_file = options.source_file
 128
 129     from aubio import onset, tempo, source, sink
 130
 131     s = source(source_file, samplerate, hopsize)
 132     if samplerate == 0: samplerate = s.get_samplerate()
 133
 134     if options.beat:
 135         o = tempo(options.onset_method, bufsize, hopsize)
 136     else:
 137         o = onset(options.onset_method, bufsize, hopsize)
 138     o.set_threshold(options.threshold)
 139
 140     timestamps = []
 141     total_frames = 0
 142     # analyze pass
 143     while True:
 144         samples, read = s()
 145         if o(samples):
 146             timestamps.append (o.get_last())
 147             if options.verbose: print "%.4f" % o.get_last_s()
 148         total_frames += read
 149         if read < hopsize: break
 150     del s
 151     # print some info
 152     nstamps = len(timestamps)
 153     duration = float (total_frames) / float(samplerate)
 154     info = 'found %(nstamps)d timestamps in %(source_file)s' % locals()
 155     info += ' (total %(duration).2fs at %(samplerate)dHz)\n' % locals()
 156     sys.stderr.write(info)
 157
 158     # cutting pass
 159     if options.cut and nstamps > 0:
 160         # generate output filenames
 161         import os
 162         source_base_name, source_ext = os.path.splitext(os.path.basename(source_file))
 163         def new_sink_name(source_base_name, timestamp):
 164             return source_base_name + '_%02.3f' % (timestamp) + '.wav'
 165         # reopen source file
 166         s = source(source_file, samplerate, hopsize)
 167         if samplerate == 0: samplerate = s.get_samplerate()
 168         # create first sink at 0
 169         g = sink(new_sink_name(source_base_name, 0.), samplerate)
 170         total_frames = 0
 171         # get next region
 172         next_onset = int(timestamps.pop(0))
 173         while True:
 174             vec, read = s()
 175             remaining = next_onset - total_frames
 176             if remaining <= read:
 177                 # write remaining samples from current region
 178                 g(vec[0:remaining], remaining)
 179                 # close this file
 180                 del g
 181                 # create a new file for the new region
 182                 g = sink(new_sink_name(source_base_name, next_onset / float(samplerate)), samplerate)
 183                 # write the remaining samples in the new file
 184                 g(vec[remaining:read], read - remaining)
 185                 #print "new slice", total_frames_written, "+", remaining, "=", start_of_next_region
 186                 if len(timestamps):
 187                     next_onset = int(timestamps.pop(0))
 188                 else:
 189                     next_onset = 1e120
 190             else:
 191                 g(vec[0:read], read)
 192             total_frames += read
 193             if read < hopsize: break
 194
 195         # print some info
 196         duration = float (total_frames) / float(samplerate)
 197         info = 'created %(nstamps)d slices from %(source_file)s' % locals()
 198         info += ' (total %(duration).2fs at %(samplerate)dHz)\n' % locals()
 199         sys.stderr.write(info)