1:mod:`audioop` --- Manipulate raw audio data
2============================================
3
4.. module:: audioop
5   :synopsis: Manipulate raw audio data.
6   :deprecated:
7
8.. deprecated-removed:: 3.11 3.13
9   The :mod:`audioop` module is deprecated
10   (see :pep:`PEP 594 <594#audioop>` for details).
11
12--------------
13
14The :mod:`audioop` module contains some useful operations on sound fragments.
15It operates on sound fragments consisting of signed integer samples 8, 16, 24
16or 32 bits wide, stored in :term:`bytes-like objects <bytes-like object>`.  All scalar items are
17integers, unless specified otherwise.
18
19.. versionchanged:: 3.4
20   Support for 24-bit samples was added.
21   All functions now accept any :term:`bytes-like object`.
22   String input now results in an immediate error.
23
24.. index::
25   single: Intel/DVI ADPCM
26   single: ADPCM, Intel/DVI
27   single: a-LAW
28   single: u-LAW
29
30This module provides support for a-LAW, u-LAW and Intel/DVI ADPCM encodings.
31
32.. This para is mostly here to provide an excuse for the index entries...
33
34A few of the more complicated operations only take 16-bit samples, otherwise the
35sample size (in bytes) is always a parameter of the operation.
36
37The module defines the following variables and functions:
38
39
40.. exception:: error
41
42   This exception is raised on all errors, such as unknown number of bytes per
43   sample, etc.
44
45
46.. function:: add(fragment1, fragment2, width)
47
48   Return a fragment which is the addition of the two samples passed as parameters.
49   *width* is the sample width in bytes, either ``1``, ``2``, ``3`` or ``4``.  Both
50   fragments should have the same length.  Samples are truncated in case of overflow.
51
52
53.. function:: adpcm2lin(adpcmfragment, width, state)
54
55   Decode an Intel/DVI ADPCM coded fragment to a linear fragment.  See the
56   description of :func:`lin2adpcm` for details on ADPCM coding. Return a tuple
57   ``(sample, newstate)`` where the sample has the width specified in *width*.
58
59
60.. function:: alaw2lin(fragment, width)
61
62   Convert sound fragments in a-LAW encoding to linearly encoded sound fragments.
63   a-LAW encoding always uses 8 bits samples, so *width* refers only to the sample
64   width of the output fragment here.
65
66
67.. function:: avg(fragment, width)
68
69   Return the average over all samples in the fragment.
70
71
72.. function:: avgpp(fragment, width)
73
74   Return the average peak-peak value over all samples in the fragment. No
75   filtering is done, so the usefulness of this routine is questionable.
76
77
78.. function:: bias(fragment, width, bias)
79
80   Return a fragment that is the original fragment with a bias added to each
81   sample.  Samples wrap around in case of overflow.
82
83
84.. function:: byteswap(fragment, width)
85
86   "Byteswap" all samples in a fragment and returns the modified fragment.
87   Converts big-endian samples to little-endian and vice versa.
88
89   .. versionadded:: 3.4
90
91
92.. function:: cross(fragment, width)
93
94   Return the number of zero crossings in the fragment passed as an argument.
95
96
97.. function:: findfactor(fragment, reference)
98
99   Return a factor *F* such that ``rms(add(fragment, mul(reference, -F)))`` is
100   minimal, i.e., return the factor with which you should multiply *reference* to
101   make it match as well as possible to *fragment*.  The fragments should both
102   contain 2-byte samples.
103
104   The time taken by this routine is proportional to ``len(fragment)``.
105
106
107.. function:: findfit(fragment, reference)
108
109   Try to match *reference* as well as possible to a portion of *fragment* (which
110   should be the longer fragment).  This is (conceptually) done by taking slices
111   out of *fragment*, using :func:`findfactor` to compute the best match, and
112   minimizing the result.  The fragments should both contain 2-byte samples.
113   Return a tuple ``(offset, factor)`` where *offset* is the (integer) offset into
114   *fragment* where the optimal match started and *factor* is the (floating-point)
115   factor as per :func:`findfactor`.
116
117
118.. function:: findmax(fragment, length)
119
120   Search *fragment* for a slice of length *length* samples (not bytes!) with
121   maximum energy, i.e., return *i* for which ``rms(fragment[i*2:(i+length)*2])``
122   is maximal.  The fragments should both contain 2-byte samples.
123
124   The routine takes time proportional to ``len(fragment)``.
125
126
127.. function:: getsample(fragment, width, index)
128
129   Return the value of sample *index* from the fragment.
130
131
132.. function:: lin2adpcm(fragment, width, state)
133
134   Convert samples to 4 bit Intel/DVI ADPCM encoding.  ADPCM coding is an adaptive
135   coding scheme, whereby each 4 bit number is the difference between one sample
136   and the next, divided by a (varying) step.  The Intel/DVI ADPCM algorithm has
137   been selected for use by the IMA, so it may well become a standard.
138
139   *state* is a tuple containing the state of the coder.  The coder returns a tuple
140   ``(adpcmfrag, newstate)``, and the *newstate* should be passed to the next call
141   of :func:`lin2adpcm`.  In the initial call, ``None`` can be passed as the state.
142   *adpcmfrag* is the ADPCM coded fragment packed 2 4-bit values per byte.
143
144
145.. function:: lin2alaw(fragment, width)
146
147   Convert samples in the audio fragment to a-LAW encoding and return this as a
148   bytes object.  a-LAW is an audio encoding format whereby you get a dynamic
149   range of about 13 bits using only 8 bit samples.  It is used by the Sun audio
150   hardware, among others.
151
152
153.. function:: lin2lin(fragment, width, newwidth)
154
155   Convert samples between 1-, 2-, 3- and 4-byte formats.
156
157   .. note::
158
159      In some audio formats, such as .WAV files, 16, 24 and 32 bit samples are
160      signed, but 8 bit samples are unsigned.  So when converting to 8 bit wide
161      samples for these formats, you need to also add 128 to the result::
162
163         new_frames = audioop.lin2lin(frames, old_width, 1)
164         new_frames = audioop.bias(new_frames, 1, 128)
165
166      The same, in reverse, has to be applied when converting from 8 to 16, 24
167      or 32 bit width samples.
168
169
170.. function:: lin2ulaw(fragment, width)
171
172   Convert samples in the audio fragment to u-LAW encoding and return this as a
173   bytes object.  u-LAW is an audio encoding format whereby you get a dynamic
174   range of about 14 bits using only 8 bit samples.  It is used by the Sun audio
175   hardware, among others.
176
177
178.. function:: max(fragment, width)
179
180   Return the maximum of the *absolute value* of all samples in a fragment.
181
182
183.. function:: maxpp(fragment, width)
184
185   Return the maximum peak-peak value in the sound fragment.
186
187
188.. function:: minmax(fragment, width)
189
190   Return a tuple consisting of the minimum and maximum values of all samples in
191   the sound fragment.
192
193
194.. function:: mul(fragment, width, factor)
195
196   Return a fragment that has all samples in the original fragment multiplied by
197   the floating-point value *factor*.  Samples are truncated in case of overflow.
198
199
200.. function:: ratecv(fragment, width, nchannels, inrate, outrate, state[, weightA[, weightB]])
201
202   Convert the frame rate of the input fragment.
203
204   *state* is a tuple containing the state of the converter.  The converter returns
205   a tuple ``(newfragment, newstate)``, and *newstate* should be passed to the next
206   call of :func:`ratecv`.  The initial call should pass ``None`` as the state.
207
208   The *weightA* and *weightB* arguments are parameters for a simple digital filter
209   and default to ``1`` and ``0`` respectively.
210
211
212.. function:: reverse(fragment, width)
213
214   Reverse the samples in a fragment and returns the modified fragment.
215
216
217.. function:: rms(fragment, width)
218
219   Return the root-mean-square of the fragment, i.e. ``sqrt(sum(S_i^2)/n)``.
220
221   This is a measure of the power in an audio signal.
222
223
224.. function:: tomono(fragment, width, lfactor, rfactor)
225
226   Convert a stereo fragment to a mono fragment.  The left channel is multiplied by
227   *lfactor* and the right channel by *rfactor* before adding the two channels to
228   give a mono signal.
229
230
231.. function:: tostereo(fragment, width, lfactor, rfactor)
232
233   Generate a stereo fragment from a mono fragment.  Each pair of samples in the
234   stereo fragment are computed from the mono sample, whereby left channel samples
235   are multiplied by *lfactor* and right channel samples by *rfactor*.
236
237
238.. function:: ulaw2lin(fragment, width)
239
240   Convert sound fragments in u-LAW encoding to linearly encoded sound fragments.
241   u-LAW encoding always uses 8 bits samples, so *width* refers only to the sample
242   width of the output fragment here.
243
244Note that operations such as :func:`.mul` or :func:`.max` make no distinction
245between mono and stereo fragments, i.e. all samples are treated equal.  If this
246is a problem the stereo fragment should be split into two mono fragments first
247and recombined later.  Here is an example of how to do that::
248
249   def mul_stereo(sample, width, lfactor, rfactor):
250       lsample = audioop.tomono(sample, width, 1, 0)
251       rsample = audioop.tomono(sample, width, 0, 1)
252       lsample = audioop.mul(lsample, width, lfactor)
253       rsample = audioop.mul(rsample, width, rfactor)
254       lsample = audioop.tostereo(lsample, width, 1, 0)
255       rsample = audioop.tostereo(rsample, width, 0, 1)
256       return audioop.add(lsample, rsample, width)
257
258If you use the ADPCM coder to build network packets and you want your protocol
259to be stateless (i.e. to be able to tolerate packet loss) you should not only
260transmit the data but also the state.  Note that you should send the *initial*
261state (the one you passed to :func:`lin2adpcm`) along to the decoder, not the
262final state (as returned by the coder).  If you want to use
263:class:`struct.Struct` to store the state in binary you can code the first
264element (the predicted value) in 16 bits and the second (the delta index) in 8.
265
266The ADPCM coders have never been tried against other ADPCM coders, only against
267themselves.  It could well be that I misinterpreted the standards in which case
268they will not be interoperable with the respective standards.
269
270The :func:`find\*` routines might look a bit funny at first sight. They are
271primarily meant to do echo cancellation.  A reasonably fast way to do this is to
272pick the most energetic piece of the output sample, locate that in the input
273sample and subtract the whole output sample from the input sample::
274
275   def echocancel(outputdata, inputdata):
276       pos = audioop.findmax(outputdata, 800)    # one tenth second
277       out_test = outputdata[pos*2:]
278       in_test = inputdata[pos*2:]
279       ipos, factor = audioop.findfit(in_test, out_test)
280       # Optional (for better cancellation):
281       # factor = audioop.findfactor(in_test[ipos*2:ipos*2+len(out_test)],
282       #              out_test)
283       prefill = '\0'*(pos+ipos)*2
284       postfill = '\0'*(len(inputdata)-len(prefill)-len(outputdata))
285       outputdata = prefill + audioop.mul(outputdata, 2, -factor) + postfill
286       return audioop.add(inputdata, outputdata, 2)
287
288