soxexam - SoX Examples (CHEAT SHEET)

Introduction

In general, SoX will attempt to take an input sound file format and convert it into a new file format using a similar data type and sample rate. For instance, "sox monkey.au monkey.wav" would try and convert the mono 8000Hz u-law sample .au file that comes with SoX to a 8000Hz u-law .wav file.

If an output format doesn't support the same data type as the input file then SoX will generally select a default data type to save it in. You can override the default data type selection by using command line options. This is also useful for producing an output file with higher or lower precision data and/or sample rate.

Most file formats that contain headers can automatically be read in. When working with header-less file formats then a user must manually tell SoX the data type and sample rate using command line options.

When working with header-less files (raw files), you may take advantage of the pseudo-file types of .ub, .uw, .sb, .sw, .ul, and .sl. By using these extensions on your filenames you will not have to specify the corresponding options on the command line.

Precision

The following data types and formats can be represented by their total uncompressed bit precision. When converting from one data type to another care must be taken to insure it has an equal or greater precision. If not then the audio quality will be degraded. This is not always a bad thing when your working with things such as voice audio and are concerned about disk space or bandwidth of the audio data.

Data Format Precision

___________ _________

unsigned byte 8-bit

signed byte 8-bit

u-law 14-bit

A-law 13-bit

unsigned word 16-bit

signed word 16-bit

ADPCM 16-bit

GSM 16-bit

unsigned long 32-bit

signed long 32-bit

___________ _________

Examples

Use the '-V' option on all your command lines. It makes SoX print out its idea of what is going on. '-V' is your friend.

To convert from unsigned bytes at 8000 Hz to signed words at 8000 Hz:

sox -r 8000 -c 1 filename.ub newfile.sw

To convert from Apple's AIFF format to Microsoft's WAV format:

sox filename.aiff filename.wav

To convert from mono raw 8000 Hz 8-bit unsigned PCM data to a WAV file:

sox -r 8000 -u -b -c 1 filename.raw filename.wav

SoX may even be used to convert sample rates. Downconverting will reduce the bandwidth of a sample, but will reduce storage space on your disk. All such conversions are lossy and will introduce some noise. You should really pass your sample through a low pass filter prior to downconverting as this will prevent alias signals (which would sound like additional noise). For example to convert from a sample recorded at 11025 Hz to a u-law file at 8000 Hz sample rate:

sox infile.wav -t au -r 8000 -U -b -c 1 outputfile.au

To add a low-pass filter (note use of stdout for output of the first stage and stdin for input on the second stage):

sox infile.wav -t raw -s -w -c 1 - lowpass 3700 | sox -t raw -r 11025 -s -w -c 1 - -t au -r 8000 -U -b -c 1 ofile.au

If you hear some clicks and pops when converting to u-law or A-law, reduce the output level slightly, for example this will decrease it by 20%:

sox infile.wav -t au -r 8000 -U -b -c 1 -v .8 outputfile.au

SoX is great to use along with other command line programs by passing data between the programs using pipelines. The most common example is to use mpg123 to convert mp3 files in to wav files. The following command line will do this:

mpg123 -b 10000 -s filename.mp3 | sox -t raw -r 44100 -s -w -c 2 - filename.wav

When working with totally unknown audio data then the "auto" file format may be of use. It attempts to guess what the file type is and then you may save it into a known audio format.

sox -V -t auto filename.snd filename.wav

It is important to understand how the internals of SoX work with compressed audio including u-law, A-law, ADPCM, or GSM. SoX takes ALL input data types and converts them to uncompressed 32-bit signed data. It will then convert this internal version into the requested output format. This means additional noise can be introduced from decompressing data and then recompressing. If applying multiple effects to audio data, it is best to save the intermediate data as PCM data. After the final effect is performed, then you can specify it as a compressed output format. This will keep noise introduction to a minimum.

The following example applies various effects to an 8000 Hz ADPCM input file and then end up with the final file as 44100 Hz ADPCM.

sox firstfile.wav -r 44100 -s -w secondfile.wav

sox secondfile.wav thirdfile.wav swap

sox thirdfile.wav -a -b finalfile.wav mask

Under a DOS shell, you can convert several audio files to an new output format using something similar to the following command line:

FOR %X IN (*.RAW) DO sox -r 11025 -w -s -t raw $X $X.wav

Special thanks goes to Juergen Mueller (jmeuller@uia.au.ac.be) for this write up on effects.

Introduction:

The core problem is that you need some experience in using effects in order to say "that any old sound file sounds with effects absolutely hip". There isn't any rule-based system which tell you the correct setting of all the parameters for every effect. But after some time you will become an expert in using effects.

Here are some examples which can be used with any music sample. (For a sample where only a single instrument is playing, extreme parameter setting may make well-known "typically" or "classical" sounds. Likewise, for drums, vocals or guitars.)

Single effects will be explained and some given parameter settings that can be used to understand the theory by listening to the sound file with the added effect.

Using multiple effects in parallel or in series can result either in a very nice sound or (mostly) in a dramatic overloading in variations of sounds such that your ear may follow the sound but you will feel unsatisfied. Hence, for the first time using effects try to compose them as minimally as possible. We don't regard the composition of effects in the examples because too many combinations are possible and you really need a very fast machine and a lot of memory to play them in real-time.

However, real-time playing of sounds will greatly speed up learning and/or tuning the parameter settings for your sounds in order to get that "perfect" effect.

Basically, we will use the "play" front-end of SoX since it is easier to listen sounds coming out of the speaker or earphone instead of looking at cryptic data in sound files.

For easy listening of file.xxx ("xxx" is any sound format):

play file.xxx effect-name effect-parameters

Or more SoX-like (for "dsp" output on a UNIX/Linux computer):

sox file.xxx -t ossdsp -w -s /dev/dsp effect-name effect-parameters

or (for "au" output):

sox file.xxx -t sunau -w -s /dev/audio effect-name effect-parameters

And for date freaks:

sox file.xxx file.yyy effect-name effect-parameters

Additional options can be used. However, in this case, for real-time playing you'll need a very fast machine.

Notes:

I played all examples in real-time on a Pentium 100 with 32 MB and Linux 2.0.30 using a self-recorded sample ( 3:15 min long in "wav" format with 44.1 kHz sample rate and stereo 16 bit ). The sample should not contain any of the effects. However, if you take any recording of a sound track from radio or tape or CD, and it sounds like a live concert or ten people are playing the same rhythm with their drums or funky-grooves, then take any other sample. (Typically, less then four different instruments and no synthesizer in the sample is suitable. Likewise, the combination vocal, drums, bass and guitar.)

Effects:

Echo

An echo effect can be naturally found in the mountains, standing somewhere on a mountain and shouting a single word will result in one or more repetitions of the word (if not, turn a bit around and try again, or climb to the next mountain).

However, the time difference between shouting and repeating is the delay (time), its loudness is the decay. Multiple echos can have different delays and decays.

It is very popular to use echos to play an instrument with itself together, like some guitar players (Brain May from Queen) or vocalists are doing. For music samples of more than one instrument, echo can be used to add a second sample shortly after the original one.

This will sound as if you are doubling the number of instruments playing in the same sample:

play file.xxx echo 0.8 0.88 60.0 0.4

If the delay is very short, then it sound like a (metallic) robot playing music:

play file.xxx echo 0.8 0.88 6.0 0.4

Longer delay will sound like an open air concert in the mountains:

play file.xxx echo 0.8 0.9 1000.0 0.3

One mountain more, and:

play file.xxx echo 0.8 0.9 1000.0 0.3 1800.0 0.25

Echos

Like the echo effect, echos stand for "ECHO in Sequel", that is the first echos takes the input, the second the input and the first echos, the third the input and the first and the second echos, ... and so on. Care should be taken using many echos (see introduction); a single echos has the same effect as a single echo.

The sample will be bounced twice in symmetric echos:

play file.xxx echos 0.8 0.7 700.0 0.25 700.0 0.3

The sample will be bounced twice in asymmetric echos:

play file.xxx echos 0.8 0.7 700.0 0.25 900.0 0.3

The sample will sound as if played in a garage:

play file.xxx echos 0.8 0.7 40.0 0.25 63.0 0.3

Chorus

The chorus effect has its name because it will often be used to make a single vocal sound like a chorus. But it can be applied to other instrument samples too.

It works like the echo effect with a short delay, but the delay isn't constant. The delay is varied using a sinusoidal or triangular modulation. The modulation depth defines the range the modulated delay is played before or after the delay. Hence the delayed sound will sound slower or faster, that is the delayed sound tuned around the original one, like in a chorus where some vocals are a bit out of tune.

The typical delay is around 40ms to 60ms, the speed of the modulation is best near 0.25Hz and the modulation depth around 2ms.

A single delay will make the sample more overloaded:

play file.xxx chorus 0.7 0.9 55.0 0.4 0.25 2.0 -t

Two delays of the original samples sound like this:

play file.xxx chorus 0.6 0.9 50.0 0.4 0.25 2.0 -t 60.0 0.32 0.4 1.3 -s

A big chorus of the sample is (three additional samples):

play file.xxx chorus 0.5 0.9 50.0 0.4 0.25 2.0 -t 60.0 0.32 0.4 2.3 -t 40.0 0.3 0.3 1.3 -s

Flanger

The flanger effect is like the chorus effect, but the delay varies between 0ms and maximal 5ms. It sound like wind blowing, sometimes faster or slower including changes of the speed.

The flanger effect is widely used in funk and soul music, where the guitar sound varies frequently slow or a bit faster.

The typical delay is around 3ms to 5ms, the speed of the modulation is best near 0.5Hz.

Now, let's groove the sample:

play file.xxx flanger 0.6 0.87 3.0 0.9 0.5 -s

listen carefully between the difference of sinusoidal and triangular modulation:

play file.xxx flanger 0.6 0.87 3.0 0.9 0.5 -t

If the decay is a bit lower, than the effect sounds more popular:

play file.xxx flanger 0.8 0.88 3.0 0.4 0.5 -t

The drunken loudspeaker system:

play file.xxx flanger 0.9 0.9 4.0 0.23 1.3 -s

Reverb

The reverb effect is often used in audience hall which are to small or contain too many many visitors which disturb (dampen) the reflection of sound at the walls. Reverb will make the sound be perceived as if it were in a large hall. You can try the reverb effect in your bathroom or garage or sport halls by shouting loud some words. You'll hear the words reflected from the walls.

The biggest problem in using the reverb effect is the correct setting of the (wall) delays such that the sound is realistic and doesn't sound like music playing in a tin can or has overloaded feedback which destroys any illusion of playing in a big hall. To help you obtain realistic reverb effects, you should decide first how long the reverb should take place until it is not loud enough to be registered by your ears. This is be done by varying the reverb time "t". To simulate small halls, use 200ms. To simulate large halls, use 1000ms. Clearly, the walls of such a hall aren't far away, so you should define its setting be given every wall its delay time. However, if the wall is to far away for the reverb time, you won't hear the reverb, so the nearest wall will be best at "t/4" delay and the farthest at "t/2". You can try other distances as well, but it won't sound very realistic. The walls shouldn't stand to close to each other and not in a multiple integer distance to each other ( so avoid wall like: 200.0 and 202.0, or something like 100.0 and 200.0 ).

Since audience halls do have a lot of walls, we will start designing one beginning with one wall:

play file.xxx reverb 1.0 600.0 180.0

One wall more:

play file.xxx reverb 1.0 600.0 180.0 200.0

Next two walls:

play file.xxx reverb 1.0 600.0 180.0 200.0 220.0 240.0

Now, why not a futuristic hall with six walls:

play file.xxx reverb 1.0 600.0 180.0 200.0 220.0 240.0 280.0 300.0

If you run out of machine power or memory, then stop as many applications as possible (every interrupt will consume a lot of CPU time which for bigger halls is absolutely necessary).

Phaser

The phaser effect is like the flanger effect, but it uses a reverb instead of an echo and does phase shifting. You'll hear the difference in the examples comparing both effects (simply change the effect name). The delay modulation can be sinusoidal or triangular, preferable is the later for multiple instruments. For single instrument sounds, the sinusoidal phaser effect will give a sharper phasing effect. The decay shouldn't be to close to 1.0 which will cause dramatic feedback. A good range is about 0.5 to 0.1 for the decay.

We will take a parameter setting as for the flanger before (gain-out is lower since feedback can raise the output dramatically):

play file.xxx phaser 0.8 0.74 3.0 0.4 0.5 -t

The drunken loudspeaker system (now less alcohol):

play file.xxx phaser 0.9 0.85 4.0 0.23 1.3 -s

A popular sound of the sample is as follows:

play file.xxx phaser 0.89 0.85 1.0 0.24 2.0 -t

The sample sounds if ten springs are in your ears:

play file.xxx phaser 0.6 0.66 3.0 0.6 2.0 -t

Compander

The compander effect allows the dynamic range of a signal to be compressed or expanded. For most situations, the attack time (response to the music getting louder) should be shorter than the decay time because our ears are more sensitive to suddenly loud music than to suddenly soft music.

For example, suppose you are listening to Strauss' "Also Sprach Zarathustra" in a noisy environment such as a car. If you turn up the volume enough to hear the soft passages over the road noise, the loud sections will be too loud. You could try this:

play file.xxx compand 0.3,1 -90,-90,-70,-70,-60,-20,0,0 -5 0 0.2

The transfer function ("-90,...") says that very soft sounds between -90 and -70 decibels (-90 is about the limit of 16-bit encoding) will remain unchanged. That keeps the compander from boosting the volume on "silent" passages such as between movements. However, sounds in the range -60 decibels to 0 decibels (maximum volume) will be boosted so that the 60-dB dynamic range of the original music will be compressed 3-to-1 into a 20-dB range, which is wide enough to enjoy the music but narrow enough to get around the road noise. The -5 dB output gain is needed to avoid clipping (the number is inexact, and was derived by experimentation). The 0 for the initial volume will work fine for a clip that starts with a bit of silence, and the delay of 0.2 has the effect of causing the compander to react a bit more quickly to sudden volume changes.

Changing the Rate of Playback

You can use stretch to change the rate of playback of an audio sample while preserving the pitch. For example to play at 1/2 the speed:

play file.wav stretch 2

To play a file at twice the speed:

play file.wav stretch .5

Other related options are "speed" to change the speed of play (and changing the pitch accordingly), and pitch, to alter the pitch of a sample. For example to speed a sample so it plays in 1/2 the time (for those Mickey Mouse voices):

play file.wav speed 2

To raise the pitch of a sample 1 while note (100 cents):

play file.wav pitch 100

Reducing noise in a recording

First find a period of silence in your recording, such as the beginning or end of a piece. If the first 1.5 seconds of the recording are silent, do

sox file.wav -t nul /dev/null trim 0 1.5 noiseprof /tmp/profile

Next, use the noisered effect to actually reduce the noise:

play file.wav noisered /tmp/profile

Other effects (copy, rate, avg, stat, vibro, lowp, highp, band, reverb)

The other effects are simple to use. However, an "easy to use manual" should be given here.

More effects (to do !)

There are a lot of effects around like noise gates, compressors, waw-waw, stereo effects and so on. They should be implemented, making SoX more useful in sound mixing techniques coming together with a great variety of different sound effects.

Combining effects by using them in parallel or serially on different channels needs some easy mechanism which is stable for use in real-time.

Really missing are the the changing of the parameters and starting/stopping of effects while playing samples in real-time!

Good luck and have fun with all the effects!

Juergen Mueller (jmueller@uia.ua.ac.be)

sox(1), play(1), rec(1)

Juergen Mueller (jmueller@uia.ua.ac.be)

Updates by Anonymous.