2 # -*- coding: utf-8 -*-
4 """aubio command line tool
6 This file was written by Paul Brossier <piem@aubio.org> and is released under
9 Note: this script is mostly about parsing command line arguments. For more
10 readable code examples, check out the `python/demos` folder."""
17 epilog = 'use "%(prog)s <command> --help" for more info about each command'
18 parser = argparse.ArgumentParser(epilog=epilog)
19 parser.add_argument('-V', '--version', help="show version",
20 action="store_true", dest="show_version")
22 subparsers = parser.add_subparsers(title='commands', dest='command',
23 parser_class= AubioArgumentParser,
26 parser_add_subcommand_help(subparsers)
28 parser_add_subcommand_onset(subparsers)
29 parser_add_subcommand_pitch(subparsers)
30 parser_add_subcommand_beat(subparsers)
31 parser_add_subcommand_tempo(subparsers)
32 parser_add_subcommand_notes(subparsers)
33 parser_add_subcommand_mfcc(subparsers)
34 parser_add_subcommand_melbands(subparsers)
35 parser_add_subcommand_quiet(subparsers)
36 parser_add_subcommand_cut(subparsers)
40 def parser_add_subcommand_help(subparsers):
41 # global help subcommand
42 subparsers.add_parser('help',
43 help='show help message',
44 formatter_class = argparse.ArgumentDefaultsHelpFormatter)
46 def parser_add_subcommand_onset(subparsers):
48 subparser = subparsers.add_parser('onset',
49 help='estimate time of onsets (beginning of sound event)',
50 formatter_class = argparse.ArgumentDefaultsHelpFormatter)
52 subparser.add_buf_hop_size()
53 helpstr = "onset novelty function"
54 helpstr += " <default|energy|hfc|complex|phase|specdiff|kl|mkl|specflux>"
55 subparser.add_method(helpstr=helpstr)
56 subparser.add_threshold()
57 subparser.add_silence()
58 subparser.add_minioi()
59 subparser.add_time_format()
60 subparser.add_verbose_help()
61 subparser.set_defaults(process=process_onset)
63 def parser_add_subcommand_pitch(subparsers):
65 subparser = subparsers.add_parser('pitch',
66 help='estimate fundamental frequency (monophonic)')
68 subparser.add_buf_hop_size(buf_size=2048)
69 helpstr = "pitch detection method <default|yinfft|yin|mcomb|fcomb|schmitt>"
70 subparser.add_method(helpstr=helpstr)
71 subparser.add_threshold()
72 subparser.add_pitch_unit()
73 subparser.add_silence()
74 subparser.add_time_format()
75 subparser.add_verbose_help()
76 subparser.set_defaults(process=process_pitch)
78 def parser_add_subcommand_beat(subparsers):
80 subparser = subparsers.add_parser('beat',
81 help='estimate location of beats')
83 subparser.add_buf_hop_size(buf_size=1024, hop_size=512)
84 subparser.add_time_format()
85 subparser.add_verbose_help()
86 subparser.set_defaults(process=process_beat)
88 def parser_add_subcommand_tempo(subparsers):
90 subparser = subparsers.add_parser('tempo',
91 help='estimate overall tempo in bpm')
93 subparser.add_buf_hop_size(buf_size=1024, hop_size=512)
94 subparser.add_time_format()
95 subparser.add_verbose_help()
96 subparser.set_defaults(process=process_tempo)
98 def parser_add_subcommand_notes(subparsers):
100 subparser = subparsers.add_parser('notes',
101 help='estimate midi-like notes (monophonic)')
102 subparser.add_input()
103 subparser.add_buf_hop_size()
104 subparser.add_silence()
105 subparser.add_release_drop()
106 subparser.add_time_format()
107 subparser.add_verbose_help()
108 subparser.set_defaults(process=process_notes)
110 def parser_add_subcommand_mfcc(subparsers):
112 subparser = subparsers.add_parser('mfcc',
113 help='extract Mel-Frequency Cepstrum Coefficients')
114 subparser.add_input()
115 subparser.add_buf_hop_size()
116 subparser.add_time_format()
117 subparser.add_verbose_help()
118 subparser.set_defaults(process=process_mfcc)
120 def parser_add_subcommand_melbands(subparsers):
121 # melbands subcommand
122 subparser = subparsers.add_parser('melbands',
123 help='extract energies in Mel-frequency bands')
124 subparser.add_input()
125 subparser.add_buf_hop_size()
126 subparser.add_time_format()
127 subparser.add_verbose_help()
128 subparser.set_defaults(process=process_melbands)
130 def parser_add_subcommand_quiet(subparsers):
132 subparser = subparsers.add_parser('quiet',
133 help='extract timestamps of quiet and loud regions')
134 subparser.add_input()
135 subparser.add_hop_size()
136 subparser.add_silence()
137 subparser.add_time_format()
138 subparser.add_verbose_help()
139 subparser.set_defaults(process=process_quiet)
141 def parser_add_subcommand_cut(subparsers):
143 subparser = subparsers.add_parser('cut',
144 help='slice at timestamps')
145 subparser.add_input()
146 helpstr = "onset novelty function"
147 helpstr += " <default|energy|hfc|complex|phase|specdiff|kl|mkl|specflux>"
148 subparser.add_method(helpstr=helpstr)
149 subparser.add_buf_hop_size()
150 subparser.add_silence()
151 subparser.add_threshold(default=0.3)
152 subparser.add_minioi()
153 subparser.add_slicer_options()
154 subparser.add_time_format()
155 subparser.add_verbose_help()
156 subparser.set_defaults(process=process_cut)
158 class AubioArgumentParser(argparse.ArgumentParser):
161 self.add_argument("source_uri", default=None, nargs='?',
162 help="input sound file to analyse", metavar = "<source_uri>")
163 self.add_argument("-i", "--input", dest = "source_uri2",
164 help="input sound file to analyse", metavar = "<source_uri>")
165 self.add_argument("-r", "--samplerate",
166 metavar = "<freq>", type=int,
167 action="store", dest="samplerate", default=0,
168 help="samplerate at which the file should be represented")
170 def add_verbose_help(self):
171 self.add_argument("-v","--verbose",
172 action="count", dest="verbose", default=1,
173 help="make lots of noise [default]")
174 self.add_argument("-q","--quiet",
175 action="store_const", dest="verbose", const=0,
178 def add_buf_hop_size(self, buf_size=512, hop_size=256):
179 self.add_buf_size(buf_size=buf_size)
180 self.add_hop_size(hop_size=hop_size)
182 def add_buf_size(self, buf_size=512):
183 self.add_argument("-B","--bufsize",
184 action="store", dest="buf_size", default=buf_size,
185 metavar = "<size>", type=int,
186 help="buffer size [default=%d]" % buf_size)
188 def add_hop_size(self, hop_size=256):
189 self.add_argument("-H","--hopsize",
190 metavar = "<size>", type=int,
191 action="store", dest="hop_size", default=hop_size,
192 help="overlap size [default=%d]" % hop_size)
194 def add_method(self, method='default', helpstr='method'):
195 self.add_argument("-m","--method",
196 metavar = "<method>", type=str,
197 action="store", dest="method", default=method,
198 help="%s [default=%s]" % (helpstr, method))
200 def add_threshold(self, default=None):
201 self.add_argument("-t","--threshold",
202 metavar = "<threshold>", type=float,
203 action="store", dest="threshold", default=default,
204 help="threshold [default=%s]" % default)
206 def add_silence(self):
207 self.add_argument("-s", "--silence",
208 metavar = "<value>", type=float,
209 action="store", dest="silence", default=-70,
210 help="silence threshold")
212 def add_release_drop(self):
213 self.add_argument("-d", "--release-drop",
214 metavar = "<value>", type=float,
215 action="store", dest="release_drop", default=10,
216 help="release drop threshold")
218 def add_minioi(self, default="12ms"):
219 self.add_argument("-M", "--minioi",
220 metavar = "<value>", type=str,
221 action="store", dest="minioi", default=default,
222 help="minimum Inter-Onset Interval [default=%s]" % default)
224 def add_pitch_unit(self, default="Hz"):
225 help_str = "frequency unit, should be one of Hz, midi, bin, cent"
226 help_str += " [default=%s]" % default
227 self.add_argument("-u", "--pitch-unit",
228 metavar = "<value>", type=str,
229 action="store", dest="pitch_unit", default=default,
232 def add_time_format(self):
233 helpstr = "select time values output format (samples, ms, seconds)"
234 helpstr += " [default=seconds]"
235 self.add_argument("-T", "--time-format",
241 def add_slicer_options(self):
242 self.add_argument("-o","--output", type = str,
243 metavar = "<outputdir>",
244 action="store", dest="output_directory", default=None,
245 help="specify path where slices of the original file should be created")
246 self.add_argument("--cut-until-nsamples", type = int,
247 metavar = "<samples>",
248 action = "store", dest = "cut_until_nsamples", default = None,
249 help="how many extra samples should be added at the end of each slice")
250 self.add_argument("--cut-every-nslices", type = int,
251 metavar = "<samples>",
252 action = "store", dest = "cut_every_nslices", default = None,
253 help="how many slices should be groupped together at each cut")
254 self.add_argument("--cut-until-nslices", type = int,
255 metavar = "<slices>",
256 action = "store", dest = "cut_until_nslices", default = None,
257 help="how many extra slices should be added at the end of each slice")
261 def samples2seconds(n_frames, samplerate):
262 return "%f\t" % (n_frames / float(samplerate))
264 def samples2milliseconds(n_frames, samplerate):
265 return "%f\t" % (1000. * n_frames / float(samplerate))
267 def samples2samples(n_frames, _samplerate):
268 return "%d\t" % n_frames
271 if mode is None or mode == 'seconds' or mode == 's':
272 return samples2seconds
273 elif mode == 'ms' or mode == 'milliseconds':
274 return samples2milliseconds
275 elif mode == 'samples':
276 return samples2samples
278 raise ValueError("invalid time format '%s'" % mode)
280 # definition of processing classes
282 class default_process(object):
283 def __init__(self, args):
284 if 'time_format' in args:
285 self.time2string = timefunc(args.time_format)
286 if args.verbose > 2 and hasattr(self, 'options'):
287 name = type(self).__name__.split('_')[1]
288 optstr = ' '.join(['running', name, 'with options', repr(self.options), '\n'])
289 sys.stderr.write(optstr)
290 def flush(self, frames_read, samplerate):
291 # optionally called at the end of process
294 def parse_options(self, args, valid_opts):
295 # get any valid options found in a dictionnary of arguments
296 options = {k :v for k,v in vars(args).items() if k in valid_opts}
297 self.options = options
299 def remap_pvoc_options(self, options):
300 # FIXME: we need to remap buf_size to win_s, hop_size to hop_s
301 # adjust python/ext/py-phasevoc.c to understand buf_size/hop_size
302 if 'buf_size' in options:
303 options['win_s'] = options['buf_size']
304 del options['buf_size']
305 if 'hop_size' in options:
306 options['hop_s'] = options['hop_size']
307 del options['hop_size']
308 self.options = options
310 class process_onset(default_process):
311 valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
312 def __init__(self, args):
313 self.parse_options(args, self.valid_opts)
314 self.onset = aubio.onset(**self.options)
315 if args.threshold is not None:
316 self.onset.set_threshold(args.threshold)
318 if args.minioi.endswith('ms'):
319 self.onset.set_minioi_ms(float(args.minioi[:-2]))
320 elif args.minioi.endswith('s'):
321 self.onset.set_minioi_s(float(args.minioi[:-1]))
323 self.onset.set_minioi(int(args.minioi))
325 self.onset.set_silence(args.silence)
326 super(process_onset, self).__init__(args)
327 def __call__(self, block):
328 return self.onset(block)
329 def repr_res(self, res, _frames_read, samplerate):
331 outstr = self.time2string(self.onset.get_last(), samplerate)
332 sys.stdout.write(outstr + '\n')
334 class process_pitch(default_process):
335 valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
336 def __init__(self, args):
337 self.parse_options(args, self.valid_opts)
338 self.pitch = aubio.pitch(**self.options)
339 if args.pitch_unit is not None:
340 self.pitch.set_unit(args.pitch_unit)
341 if args.threshold is not None:
342 self.pitch.set_tolerance(args.threshold)
343 if args.silence is not None:
344 self.pitch.set_silence(args.silence)
345 super(process_pitch, self).__init__(args)
346 def __call__(self, block):
347 return self.pitch(block)
348 def repr_res(self, res, frames_read, samplerate):
349 fmt_out = self.time2string(frames_read, samplerate)
350 sys.stdout.write(fmt_out + "%.6f\n" % res[0])
352 class process_beat(default_process):
353 valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
354 def __init__(self, args):
355 self.parse_options(args, self.valid_opts)
356 self.tempo = aubio.tempo(**self.options)
357 super(process_beat, self).__init__(args)
358 def __call__(self, block):
359 return self.tempo(block)
360 def repr_res(self, res, _frames_read, samplerate):
362 outstr = self.time2string(self.tempo.get_last(), samplerate)
363 sys.stdout.write(outstr + '\n')
365 class process_tempo(process_beat):
366 def __init__(self, args):
367 super(process_tempo, self).__init__(args)
368 self.beat_locations = []
369 def repr_res(self, res, _frames_read, samplerate):
371 self.beat_locations.append(self.tempo.get_last_s())
372 def flush(self, frames_read, samplerate):
374 if len(self.beat_locations) < 2:
375 outstr = "unknown bpm"
377 bpms = 60./ np.diff(self.beat_locations)
378 median_bpm = np.mean(bpms)
379 if len(self.beat_locations) < 10:
380 outstr = "%.2f bpm (uncertain)" % median_bpm
382 outstr = "%.2f bpm" % median_bpm
383 sys.stdout.write(outstr + '\n')
385 class process_notes(default_process):
386 valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
387 def __init__(self, args):
388 self.parse_options(args, self.valid_opts)
389 self.notes = aubio.notes(**self.options)
390 if args.silence is not None:
391 self.notes.set_silence(args.silence)
392 if args.release_drop is not None:
393 self.notes.set_release_drop(args.release_drop)
394 super(process_notes, self).__init__(args)
395 def __call__(self, block):
396 return self.notes(block)
397 def repr_res(self, res, frames_read, samplerate):
398 if res[2] != 0: # note off
399 fmt_out = self.time2string(frames_read, samplerate)
400 sys.stdout.write(fmt_out + '\n')
401 if res[0] != 0: # note on
403 fmt_out = "%f\t" % lastmidi
404 fmt_out += self.time2string(frames_read, samplerate)
405 sys.stdout.write(fmt_out) # + '\t')
406 def flush(self, frames_read, samplerate):
407 eof = self.time2string(frames_read, samplerate)
408 sys.stdout.write(eof + '\n')
410 class process_mfcc(default_process):
411 def __init__(self, args):
412 valid_opts1 = ['hop_size', 'buf_size']
413 self.parse_options(args, valid_opts1)
414 self.remap_pvoc_options(self.options)
415 self.pv = aubio.pvoc(**self.options)
417 valid_opts2 = ['buf_size', 'n_filters', 'n_coeffs', 'samplerate']
418 self.parse_options(args, valid_opts2)
419 self.mfcc = aubio.mfcc(**self.options)
421 # remember all options
422 self.parse_options(args, list(set(valid_opts1 + valid_opts2)))
424 super(process_mfcc, self).__init__(args)
426 def __call__(self, block):
427 fftgrain = self.pv(block)
428 return self.mfcc(fftgrain)
429 def repr_res(self, res, frames_read, samplerate):
430 fmt_out = self.time2string(frames_read, samplerate)
431 fmt_out += ' '.join(["% 9.7f" % f for f in res.tolist()])
432 sys.stdout.write(fmt_out + '\n')
434 class process_melbands(default_process):
435 def __init__(self, args):
437 valid_opts = ['hop_size', 'buf_size']
438 self.parse_options(args, valid_opts)
439 self.remap_pvoc_options(self.options)
440 self.pv = aubio.pvoc(**self.options)
442 valid_opts = ['buf_size', 'n_filters']
443 self.parse_options(args, valid_opts)
444 self.remap_pvoc_options(self.options)
445 self.filterbank = aubio.filterbank(**self.options)
446 self.filterbank.set_mel_coeffs_slaney(args.samplerate)
448 super(process_melbands, self).__init__(args)
449 def __call__(self, block):
450 fftgrain = self.pv(block)
451 return self.filterbank(fftgrain)
452 def repr_res(self, res, frames_read, samplerate):
453 fmt_out = self.time2string(frames_read, samplerate)
454 fmt_out += ' '.join(["% 9.7f" % f for f in res.tolist()])
455 sys.stdout.write(fmt_out + '\n')
457 class process_quiet(default_process):
458 def __init__(self, args):
460 valid_opts = ['hop_size', 'silence']
461 self.parse_options(args, valid_opts)
464 if args.silence is not None:
465 self.silence = args.silence
466 super(process_quiet, self).__init__(args)
468 def __call__(self, block):
469 if aubio.silence_detection(block, self.silence) == 1:
470 if self.wassilence != 1:
472 return 2 # newly found silence
473 return 1 # silence again
475 if self.wassilence != 0:
477 return -1 # newly found noise
478 return 0 # noise again
480 def repr_res(self, res, frames_read, samplerate):
486 if fmt_out is not None:
487 fmt_out += self.time2string(frames_read, samplerate)
488 sys.stdout.write(fmt_out + '\n')
490 class process_cut(process_onset):
491 def __init__(self, args):
492 super(process_cut, self).__init__(args)
496 def __call__(self, block):
497 ret = super(process_cut, self).__call__(block)
498 if ret: self.slices.append(self.onset.get_last())
501 def flush(self, frames_read, samplerate):
502 from aubio.cut import _cut_slice
503 _cut_slice(self.options, self.slices)
504 duration = float (frames_read) / float(samplerate)
505 base_info = '%(source_file)s' % {'source_file': self.options.source_uri}
506 base_info += ' (total %(duration).2fs at %(samplerate)dHz)\n' % \
507 {'duration': duration, 'samplerate': samplerate}
508 info = "created %d slices from " % len(self.slices)
510 sys.stderr.write(info)
513 parser = aubio_parser()
514 args = parser.parse_args()
515 if 'show_version' in args and args.show_version:
516 sys.stdout.write('aubio version ' + aubio.version + '\n')
518 elif 'verbose' in args and args.verbose > 3:
519 sys.stderr.write('aubio version ' + aubio.version + '\n')
520 if 'command' not in args or args.command is None or args.command in ['help']:
521 # no command given, print help and return 1
523 if args.command and args.command in ['help']:
527 elif not args.source_uri and not args.source_uri2:
528 sys.stderr.write("Error: a source is required\n")
531 elif args.source_uri2 is not None:
532 args.source_uri = args.source_uri2
535 with aubio.source(args.source_uri, hop_size=args.hop_size,
536 samplerate=args.samplerate) as a_source:
537 # always update args.samplerate to native samplerate, in case
538 # source was opened with args.samplerate=0
539 args.samplerate = a_source.samplerate
540 # create the processor for this subcommand
541 processor = args.process(args)
544 # read new block from source
545 block, read = a_source()
546 # execute processor on this block
547 res = processor(block)
548 # print results for this block
550 processor.repr_res(res, frames_read, a_source.samplerate)
551 # increment total number of frames read
553 # exit loop at end of file
554 if read < a_source.hop_size: break
555 # flush the processor if needed
556 processor.flush(frames_read, a_source.samplerate)
558 fmt_string = "read {:.2f}s"
559 fmt_string += " ({:d} samples in {:d} blocks of {:d})"
560 fmt_string += " from {:s} at {:d}Hz\n"
561 sys.stderr.write(fmt_string.format(
562 frames_read/float(a_source.samplerate),
564 frames_read // a_source.hop_size + 1,
567 a_source.samplerate))
568 except KeyboardInterrupt: