1:mod:`audioop` --- Manipulate raw audio data
2============================================
3
4.. module:: audioop
5   :synopsis: Manipulate raw audio data.
6
7--------------
8
9The :mod:`audioop` module contains some useful operations on sound fragments.
10It operates on sound fragments consisting of signed integer samples 8, 16, 24
11or 32 bits wide, stored in :term:`bytes-like objects <bytes-like object>`.  All scalar items are
12integers, unless specified otherwise.
13
14.. versionchanged:: 3.4
15   Support for 24-bit samples was added.
16   All functions now accept any :term:`bytes-like object`.
17   String input now results in an immediate error.
18
19.. index::
20   single: Intel/DVI ADPCM
21   single: ADPCM, Intel/DVI
22   single: a-LAW
23   single: u-LAW
24
25This module provides support for a-LAW, u-LAW and Intel/DVI ADPCM encodings.
26
27.. This para is mostly here to provide an excuse for the index entries...
28
29A few of the more complicated operations only take 16-bit samples, otherwise the
30sample size (in bytes) is always a parameter of the operation.
31
32The module defines the following variables and functions:
33
34
35.. exception:: error
36
37   This exception is raised on all errors, such as unknown number of bytes per
38   sample, etc.
39
40
41.. function:: add(fragment1, fragment2, width)
42
43   Return a fragment which is the addition of the two samples passed as parameters.
44   *width* is the sample width in bytes, either ``1``, ``2``, ``3`` or ``4``.  Both
45   fragments should have the same length.  Samples are truncated in case of overflow.
46
47
48.. function:: adpcm2lin(adpcmfragment, width, state)
49
50   Decode an Intel/DVI ADPCM coded fragment to a linear fragment.  See the
51   description of :func:`lin2adpcm` for details on ADPCM coding. Return a tuple
52   ``(sample, newstate)`` where the sample has the width specified in *width*.
53
54
55.. function:: alaw2lin(fragment, width)
56
57   Convert sound fragments in a-LAW encoding to linearly encoded sound fragments.
58   a-LAW encoding always uses 8 bits samples, so *width* refers only to the sample
59   width of the output fragment here.
60
61
62.. function:: avg(fragment, width)
63
64   Return the average over all samples in the fragment.
65
66
67.. function:: avgpp(fragment, width)
68
69   Return the average peak-peak value over all samples in the fragment. No
70   filtering is done, so the usefulness of this routine is questionable.
71
72
73.. function:: bias(fragment, width, bias)
74
75   Return a fragment that is the original fragment with a bias added to each
76   sample.  Samples wrap around in case of overflow.
77
78
79.. function:: byteswap(fragment, width)
80
81   "Byteswap" all samples in a fragment and returns the modified fragment.
82   Converts big-endian samples to little-endian and vice versa.
83
84   .. versionadded:: 3.4
85
86
87.. function:: cross(fragment, width)
88
89   Return the number of zero crossings in the fragment passed as an argument.
90
91
92.. function:: findfactor(fragment, reference)
93
94   Return a factor *F* such that ``rms(add(fragment, mul(reference, -F)))`` is
95   minimal, i.e., return the factor with which you should multiply *reference* to
96   make it match as well as possible to *fragment*.  The fragments should both
97   contain 2-byte samples.
98
99   The time taken by this routine is proportional to ``len(fragment)``.
100
101
102.. function:: findfit(fragment, reference)
103
104   Try to match *reference* as well as possible to a portion of *fragment* (which
105   should be the longer fragment).  This is (conceptually) done by taking slices
106   out of *fragment*, using :func:`findfactor` to compute the best match, and
107   minimizing the result.  The fragments should both contain 2-byte samples.
108   Return a tuple ``(offset, factor)`` where *offset* is the (integer) offset into
109   *fragment* where the optimal match started and *factor* is the (floating-point)
110   factor as per :func:`findfactor`.
111
112
113.. function:: findmax(fragment, length)
114
115   Search *fragment* for a slice of length *length* samples (not bytes!) with
116   maximum energy, i.e., return *i* for which ``rms(fragment[i*2:(i+length)*2])``
117   is maximal.  The fragments should both contain 2-byte samples.
118
119   The routine takes time proportional to ``len(fragment)``.
120
121
122.. function:: getsample(fragment, width, index)
123
124   Return the value of sample *index* from the fragment.
125
126
127.. function:: lin2adpcm(fragment, width, state)
128
129   Convert samples to 4 bit Intel/DVI ADPCM encoding.  ADPCM coding is an adaptive
130   coding scheme, whereby each 4 bit number is the difference between one sample
131   and the next, divided by a (varying) step.  The Intel/DVI ADPCM algorithm has
132   been selected for use by the IMA, so it may well become a standard.
133
134   *state* is a tuple containing the state of the coder.  The coder returns a tuple
135   ``(adpcmfrag, newstate)``, and the *newstate* should be passed to the next call
136   of :func:`lin2adpcm`.  In the initial call, ``None`` can be passed as the state.
137   *adpcmfrag* is the ADPCM coded fragment packed 2 4-bit values per byte.
138
139
140.. function:: lin2alaw(fragment, width)
141
142   Convert samples in the audio fragment to a-LAW encoding and return this as a
143   bytes object.  a-LAW is an audio encoding format whereby you get a dynamic
144   range of about 13 bits using only 8 bit samples.  It is used by the Sun audio
145   hardware, among others.
146
147
148.. function:: lin2lin(fragment, width, newwidth)
149
150   Convert samples between 1-, 2-, 3- and 4-byte formats.
151
152   .. note::
153
154      In some audio formats, such as .WAV files, 16, 24 and 32 bit samples are
155      signed, but 8 bit samples are unsigned.  So when converting to 8 bit wide
156      samples for these formats, you need to also add 128 to the result::
157
158         new_frames = audioop.lin2lin(frames, old_width, 1)
159         new_frames = audioop.bias(new_frames, 1, 128)
160
161      The same, in reverse, has to be applied when converting from 8 to 16, 24
162      or 32 bit width samples.
163
164
165.. function:: lin2ulaw(fragment, width)
166
167   Convert samples in the audio fragment to u-LAW encoding and return this as a
168   bytes object.  u-LAW is an audio encoding format whereby you get a dynamic
169   range of about 14 bits using only 8 bit samples.  It is used by the Sun audio
170   hardware, among others.
171
172
173.. function:: max(fragment, width)
174
175   Return the maximum of the *absolute value* of all samples in a fragment.
176
177
178.. function:: maxpp(fragment, width)
179
180   Return the maximum peak-peak value in the sound fragment.
181
182
183.. function:: minmax(fragment, width)
184
185   Return a tuple consisting of the minimum and maximum values of all samples in
186   the sound fragment.
187
188
189.. function:: mul(fragment, width, factor)
190
191   Return a fragment that has all samples in the original fragment multiplied by
192   the floating-point value *factor*.  Samples are truncated in case of overflow.
193
194
195.. function:: ratecv(fragment, width, nchannels, inrate, outrate, state[, weightA[, weightB]])
196
197   Convert the frame rate of the input fragment.
198
199   *state* is a tuple containing the state of the converter.  The converter returns
200   a tuple ``(newfragment, newstate)``, and *newstate* should be passed to the next
201   call of :func:`ratecv`.  The initial call should pass ``None`` as the state.
202
203   The *weightA* and *weightB* arguments are parameters for a simple digital filter
204   and default to ``1`` and ``0`` respectively.
205
206
207.. function:: reverse(fragment, width)
208
209   Reverse the samples in a fragment and returns the modified fragment.
210
211
212.. function:: rms(fragment, width)
213
214   Return the root-mean-square of the fragment, i.e. ``sqrt(sum(S_i^2)/n)``.
215
216   This is a measure of the power in an audio signal.
217
218
219.. function:: tomono(fragment, width, lfactor, rfactor)
220
221   Convert a stereo fragment to a mono fragment.  The left channel is multiplied by
222   *lfactor* and the right channel by *rfactor* before adding the two channels to
223   give a mono signal.
224
225
226.. function:: tostereo(fragment, width, lfactor, rfactor)
227
228   Generate a stereo fragment from a mono fragment.  Each pair of samples in the
229   stereo fragment are computed from the mono sample, whereby left channel samples
230   are multiplied by *lfactor* and right channel samples by *rfactor*.
231
232
233.. function:: ulaw2lin(fragment, width)
234
235   Convert sound fragments in u-LAW encoding to linearly encoded sound fragments.
236   u-LAW encoding always uses 8 bits samples, so *width* refers only to the sample
237   width of the output fragment here.
238
239Note that operations such as :func:`.mul` or :func:`.max` make no distinction
240between mono and stereo fragments, i.e. all samples are treated equal.  If this
241is a problem the stereo fragment should be split into two mono fragments first
242and recombined later.  Here is an example of how to do that::
243
244   def mul_stereo(sample, width, lfactor, rfactor):
245       lsample = audioop.tomono(sample, width, 1, 0)
246       rsample = audioop.tomono(sample, width, 0, 1)
247       lsample = audioop.mul(lsample, width, lfactor)
248       rsample = audioop.mul(rsample, width, rfactor)
249       lsample = audioop.tostereo(lsample, width, 1, 0)
250       rsample = audioop.tostereo(rsample, width, 0, 1)
251       return audioop.add(lsample, rsample, width)
252
253If you use the ADPCM coder to build network packets and you want your protocol
254to be stateless (i.e. to be able to tolerate packet loss) you should not only
255transmit the data but also the state.  Note that you should send the *initial*
256state (the one you passed to :func:`lin2adpcm`) along to the decoder, not the
257final state (as returned by the coder).  If you want to use
258:class:`struct.Struct` to store the state in binary you can code the first
259element (the predicted value) in 16 bits and the second (the delta index) in 8.
260
261The ADPCM coders have never been tried against other ADPCM coders, only against
262themselves.  It could well be that I misinterpreted the standards in which case
263they will not be interoperable with the respective standards.
264
265The :func:`find\*` routines might look a bit funny at first sight. They are
266primarily meant to do echo cancellation.  A reasonably fast way to do this is to
267pick the most energetic piece of the output sample, locate that in the input
268sample and subtract the whole output sample from the input sample::
269
270   def echocancel(outputdata, inputdata):
271       pos = audioop.findmax(outputdata, 800)    # one tenth second
272       out_test = outputdata[pos*2:]
273       in_test = inputdata[pos*2:]
274       ipos, factor = audioop.findfit(in_test, out_test)
275       # Optional (for better cancellation):
276       # factor = audioop.findfactor(in_test[ipos*2:ipos*2+len(out_test)],
277       #              out_test)
278       prefill = '\0'*(pos+ipos)*2
279       postfill = '\0'*(len(inputdata)-len(prefill)-len(outputdata))
280       outputdata = prefill + audioop.mul(outputdata, 2, -factor) + postfill
281       return audioop.add(inputdata, outputdata, 2)
282
283