1
2Normalize
3
4 This is release 0.7.7 of Normalize, an audio file volume normalizer.
5
6 Copyright (c) 1999--2005, Chris Vaill <chrisvaill at gmail>
7
8 Normalize is a tool for adjusting the volume of audio files to a
9 standard level. This is useful for things like creating mixed CD's and
10 mp3 collections, where different recording levels on different albums
11 can cause the volume to vary greatly from song to song.
12
13 Send bug reports, suggestions, comments to chrisvaill at gmail.
14
15 normalize is free software. See the file COPYING for copying
16 conditions.
17 _________________________________________________________________
18
19Installation synopsis
20
21 1. ./configure options
22 2. make
23 3. make install
24
25 See the file INSTALL for more extensive directions. See the man page,
26 normalize.1, for usage. Run "./configure --help" for configure
27 options.
28 _________________________________________________________________
29
30Dependencies
31
32 These dependencies are optional. Normalize doesn't require any other
33 packages to compile and run.
34
35 MAD library (http://www.underbit.com/products/mad/)
36
37 Normalize will use the MAD MPEG Audio Decoder library if you have it
38 (highly recommended). This gives normalize the ability to read mp3
39 files. MAD support in normalize was developed using MAD version
40 0.14.2b; earlier versions may not work.
41
42 You can run configure with the --without-mad option to turn off mp3
43 read support.
44
45 Audiofile library (http://www.68k.org/~michael/audiofile/)
46
47 Normalize can use the audiofile library if version 0.2.2 or later is
48 available on your system. This gives normalize the ability to read and
49 write AIFF, AIFF-C, WAV, NeXT/Sun .snd/.au, Berkeley/IRCAM/CARL, and
50 whatever else the audiofile library people decide to implement in the
51 future.
52
53 Audiofile support is not turned on by default, because the built-in
54 WAV support is faster (only because it's specifically tailored for PCM
55 WAVs), and because I'm guessing most people only ever need to
56 normalize standard PCM WAV and mp3 files. If you only want to use
57 normalize on standard PCM WAV and mp3 files, you don't need audiofile.
58 If, however, you would like to be able to normalize all the different
59 audio file formats that audiofile handles, run configure with the
60 --with-audiofile option to turn on audiofile support.
61
62 XMMS (http://www.xmms.org/)
63
64 If you have xmms installed, the configure system will build the
65 xmms-rva plugin, which honors the relative volume adjustment frames
66 that normalize adds to ID3 tags. The option --disable-xmms prevents
67 the plugin from being built.
68 _________________________________________________________________
69
70Questions and Answers
71
72 1. What platforms does normalize work on?
73
74 I've tested normalize on GNU/Linux and FreeBSD on x86, Solaris on
75 Sparc, and Irix on MIPS. I've heard that it works on GNU/Linux on
76 Alpha and on BeOS R5. As far as Windows is concerned, you can compile
77 it using the Cygwin toolkit. Question 8, below, contains a brief
78 overview of this process.
79
80 I've tried to make the code as portable as possible, so I'd appreciate
81 hearing whether normalize works on other platforms.
82
83 2. What is normalize useful for?
84
85 Example 1. Let's say you've got a bunch of wav files containing what
86 are, in your estimation, Elvis's greatest hits, collected from various
87 albums. You want to encode them as mp3's and add them to an
88 established collection, but since they're all from different albums,
89 they're all recorded at different volumes from each other and from the
90 rest of your mp3 collection. If you've been using normalize on all
91 your wav files before you encode them, your collection is normalized
92 to the default volume level, and you want these new additions to be at
93 the same level. Just run normalize with no options on the files, and
94 each will be adjusted to the proper volume level:
95
96 normalize "Hound Dog.wav" "Blue Suede Shoes.wav" \
97 "Here Comes Santa Claus.wav" ...
98
99 Example 2. Suppose now you've just extracted all the wav files from
100 the Gorilla Biscuits album "Start Today," which, you may know, is
101 recorded at a particularly low volume. We want to make the whole album
102 louder, but individual tracks should stay at the same volume relative
103 to each other. For this we use batch mode. Say the files are named
104 01.wav to 14.wav, and are in the current directory. We invoke
105 normalize in batch mode to preserve the relative volumes, but
106 otherwise, everything's the default:
107
108 normalize -b *.wav
109
110 You can then fire up your mp3 encoder, and the whole album will be
111 uniformly louder.
112
113 Example 3. Now suppose we want to encode the Converge album "When
114 Forever Comes Crashing." This album has one song, "Ten Cents," that is
115 really quiet while the rest of the songs have about the same (loud)
116 volume. We'll turn up the verbosity so we can see what's going on:
117
118 > normalize -bv *.wav
119 Computing levels...
120 Level for track01.cdda.wav: -9.3980dBFS (0.0000dBFS peak)
121 Level for track02.cdda.wav: -9.2464dBFS (-0.1538dBFS peak)
122 Level for track03.cdda.wav: -8.6308dBFS (-0.2520dBFS peak)
123 Level for track04.cdda.wav: -8.7390dBFS (0.0000dBFS peak)
124 Level for track05.cdda.wav: -8.1000dBFS (-0.0003dBFS peak)
125 Level for track06.cdda.wav: -8.2215dBFS (-0.1754dBFS peak)
126 Level for track07.cdda.wav: -8.9346dBFS (-0.1765dBFS peak)
127 Level for track08.cdda.wav: -13.6175dBFS (-0.4552dBFS peak)
128 Level for track09.cdda.wav: -9.0107dBFS (-0.1778dBFS peak)
129 Level for track10.cdda.wav: -8.1824dBFS (-0.4519dBFS peak)
130 Level for track11.cdda.wav: -8.5700dBFS (-0.1778dBFS peak)
131 Standard deviation is 1.47 dB
132 Throwing out level of -13.6175dBFS (different by 4.58dB)
133 Average level: -8.6929dBFS
134 Applying adjustment of -3.35dB...
135
136 The volume of "Ten Cents," which is track 8, is 4.58 decibels off the
137 average, which, given a standard deviation of 1.47 decibels, makes it
138 a statistical aberration (which I've defined as anything off by more
139 that twice the standard deviation, but you can set a constant decibel
140 threshold with the -t option). Therefore, it isn't counted in the
141 average, and the adjustment applied to the album isn't thrown off
142 because of one song. Although the aberrant song's volume is not
143 counted in the average, it is adjusted along with the rest of the
144 files.
145
146 Example 4. Finally, say you want to make a mixed CD of 80's songs for
147 your mom or something. You won't allow any 80's songs to taint your
148 hallowed mp3 collection, so the absolute volumes of these tracks don't
149 matter, as long as they're all about the same, so mom doesn't have to
150 keep adjusting the volume. For this, use the mix mode option,
151
152 normalize -m *.wav
153
154 and each track will be adjusted to the average level of all the
155 tracks.
156
157 3. How does normalize work?
158
159 A little background on how normalize computes the volume of a wav
160 file, in case you want to know just how your files are being munged:
161
162 The volumes calculated are RMS amplitudes, which correspond (roughly)
163 to perceived volume. Taking the RMS amplitude of an entire file would
164 not give us quite the measure we want, though, because a quiet song
165 punctuated by short loud parts would average out to a quiet song, and
166 the adjustment we would compute would make the loud parts excessively
167 loud.
168
169 What we want is to consider the maximum volume of the file, and
170 normalize according to that. We break up the signal into 100 chunks
171 per second, and get the signal power of each chunk, in order to get an
172 estimation of "instantaneous power" over time. This "instantaneous
173 power" signal varies too much to get a good measure of the original
174 signal's maximum sustained power, so we run a smoothing algorithm over
175 the power signal (specifically, a mean filter with a window width of
176 100 elements). The maximum point of the smoothed power signal turns
177 out to be a good measure of the maximum sustained power of the file.
178 We can then take the square root of the power to get maximum sustained
179 RMS amplitude.
180
181 As for the default target amplitude of 0.25 (-12dBFS), I've found that
182 it's pretty close to the level of most of my albums already, but not
183 so high as to cause a lot of limiting on quieter albums. You may want
184 to choose a different target amplitude, depending on your music
185 collection (just make sure you normalize everything to the same
186 amplitude if you want it to all be the same volume!).
187
188 Regarding clipping: since version 0.6, a limiter is employed to
189 eliminate clipping. The limiter is on by default; you don't have to do
190 anything to use it. The 0.5 series had a -c option to turn on
191 limiting, but that limiter caused problems with inexact volume
192 adjustment. The new limiter doesn't have this problem, and the -c
193 option is considered deprecated (it will be removed in version 1.0).
194
195 Please note that I'm not a recording engineer or an electrical
196 engineer, so my signal processing theory may be off. I'd be glad to
197 hear from any signal processing wizards if I've made faulty
198 assumptions regarding signal power, perceived volume, or any of that
199 fun signal theory stuff.
200
201 4. Why don't you normalize using peak levels instead of RMS amplitude?
202
203 Well, in early (unreleased) versions, this is how it worked. I found
204 that this just didn't work well. The volume that your ear hears
205 corresponds more closely with average RMS amplitude level than with
206 peak level. Therefore, making the RMS amplitude of two files equal
207 makes their perceived volume equal. (Approximately equal, anyway:
208 certain frequencies sound louder at the same amplitude because the ear
209 is just more sensitive to those frequencies. I may try to take this
210 into account in a future version, but that opens up a whole new can of
211 worms.)
212
213 "Normalizing" by peak level generally makes files with small dynamic
214 range very loud and does nothing to files with large dynamic ranges.
215 There's not really any normalization being done, it's more of a
216 histogram expansion. That said, since version 0.5, you can use the
217 --peak option to do this in normalize if you're sure it's what you
218 really want to do.
219
220 5. Can normalize operate directly on mp3 files?
221
222 Version 0.7 and up can operate directly on MPEG audio files. An mp3
223 file is decoded (using Robert Leslie's MAD library) and analyzed on
224 the fly, without the need for large temporary WAV files. The mp3 file
225 is then "adjusted" by setting its relative volume adjustment
226 information (technically, an "RVA2" frame is set in its ID3v2 tag).
227 The advantage of this method is that the audio data doesn't need to be
228 touched, and you don't incur the cost of re-encoding. The disadvantage
229 is that your mp3 player needs to read and use relative volume
230 adjustment ID3 frames. The normalize distribution now includes a
231 plugin for xmms that honors volume adjustment frames. If you use an
232 mp3 player other than xmms, you'll have to bug the author to support
233 RVA2 frames in ID3 tags.
234
235 If you'd rather change the volume of the mp3 audio data itself, you
236 still have to decode to WAV, normalize the WAV, and re-encode. A
237 script, normalize-mp3, is included in the normalize distribution to do
238 this for you.
239
240 6. Can normalize operate on ogg vorbis files?
241
242 Version 0.8 will at least be able to read vorbis audio files.
243 Adjusting is harder, though: the problem is that, unlike with ID3, as
244 far as I know there's no standardized volume adjustment tag for ogg. I
245 could just use, say, "VOLUME_ADJUST=X.XXdB" as an ogg comment, but
246 there would be no reason for players to support it.
247
248 It may be possible to twiddle the vorbis data itself to alter the
249 volume in a lossless way. I'm looking into this, but it would be a big
250 undertaking, not something that would be finished anytime soon.
251
252 The current situation is that you have to decode to WAV, normalize the
253 WAV, and re-encode. The normalize-ogg script is included in the
254 normalize distribution to do this for you.
255
256 7. How do I normalize a whole tree of files recursively?
257
258 The "unix way" to do this is to use find:
259
260 find . -type d -exec sh -c "normalize -b \"{}\"/*.mp3" \;
261
262 will go directory by directory, running normalize -b on all mp3 files
263 in each. If you don't want batch mode, just:
264
265 find . -name \*.mp3 -exec normalize {} \;
266
267 will run normalize on each mp3 file separately. If you want to run
268 normalize in batch or mix mode on all files in the directory tree,
269 use:
270
271 find . -name \*.mp3 -print0 | xargs -0 normalize -b
272
273 A built-in recurse option has been a very popular request, so I'm
274 adding support for it in version 0.8.
275
276 8. How do I use normalize in Windows?
277
278 "I click on INSTALL but nothing happens. What's wrong?" Okay, here's
279 the deal: normalize is free software, written for free operating
280 systems such as Linux and FreeBSD. These happen to be unix-style
281 operating systems, so normalize generally works on other non-free
282 flavors of unix as well. Unlike Windows software, unix software such
283 as normalize is meant to run on many different operating systems on
284 many different architectures, so usually it comes in source code form
285 and you have to compile it for your particular setup. If you are
286 running some form of unix, normalize should compile right out of the
287 box (let me know if it doesn't!). For other operating systems, such as
288 Amiga, BeOS, OS/2, or Windows, you may have to jump through some hoops
289 to get it to compile.
290
291 A discussion of compiling unix software for Windows is way beyond the
292 scope of this FAQ, but here's a quick rundown:
293
294 1. You first need the Cygwin toolkit. After installing, start up a
295 cygwin bash shell.
296 2. Go to the directory where you unzipped the normalize archive -- it
297 would be named something like normalize-x.y.z.
298 3. Type "./configure", then "make", then "make install"
299 4. If there were no errors, you can run normalize by typing
300 "normalize" at the prompt. Normalize is a command-line utility, so
301 you have to pass it command line options. Run "normalize --help"
302 for a synopsis.
303 _________________________________________________________________
304
305 Copyright (c) 1999--2005, Chris Vaill <chrisvaill at gmail>
306
307 Permission is granted to copy, distribute, and/or modify this document
308 under the terms of the GNU General Public License as published by the
309 Free Software Foundation; either version 2 of the License, or (at your
310 option) any later version.
311