Real-time mp3 recording

ArticleCategory: [Artikel Kategorie]


AuthorImage:[Bild des Autors]

[Photo of the Author]

TranslationInfo:[Author and translation history]

original in nl Philip de Groot

nl to en Philip de Groot

AboutTheAuthor:[Über den Autor]

Soon, he will get his Ph.D. at the University of Nijmegen on the subject of chemometrics. He enjoys working with Linux and writes down his experiences on a regular basis.


This articles shows how to record music with Linux. In contradiction to the LinuxFocus tip of January 2001, not the 'expensive' WAV-format is used. Instead, the much smaller mp3-format is directly recorded. 24 hour recordings in CD-quality become reality when using this format.

ArticleIllustration:[Titelbild des Artikels]


ArticleBody:[Der eigentliche Artikel]


Usually, it is impossible to directly convert wave-sound to mp3-sound in real-time. The procedure that one should follow is:
  • First, record a wave-file with a recording-program.
  • Use an mp3-encoder (e.g. 'Lame') to convert the wave-file to an mp3-file.

Recording a wave-file quickly takes up quite a lot of disk space (100 MB for 10 minutes of stereo music in CD-quality). This restricts the storage capacity for music recorded in this format. An mp3-encoder is capable of converting the wave-file of 100 MB to 10 MB in the standard mp3-quality (128 kbps, stereo). This property enables the digital recording of music for 24 hours, without any interruption! 24 hours is 60*24 = 1440 minutes, using only 1,4 GB (1440 MB) hard disk space!!! I have used this feature to record the eighties top 300, transmitted by Loostad radio (some English pages are also available). The recording was interrupted just once, the mp3-recording program stops automatically after 24 hours and this was not documentated..., the recording was very good! I will show you what software I used and explain alternative ways to make real-time mp3-recordings. I have as well made the software used for this article available on seperate page (links are further down in the article).

The procedure

Below, two methods for making real-time, stereo, and CD-quality mp3-recordings are discussed:

'Mpegrec' does the same as 'sox | lame', but different settings are already defined. The user does not need anything else. Just installl 'Lame'! I advise you to use 'mpegrec'. However, keep in mind that this program stops after 24 hours!!! Furthermore, an annoying casting-bug results in erroneous recording times: the recording time as set by the user is not correctly used. Most of the times, it is quite a lot shorter!

How to use mpegrec

Using 'mpegrec' is easy. In an xterm, type 'mpegrec -b 128 -o ./test.mp3' and there you go!
Explanation of the different options:
-b Use 128 kbps as the mp3 bitrate. The standard value in 'mpegrec' is 192, which is a strange (and much too high) value. Almost all mp3-files have a constant bitrate of 128 kbps.
-o Name of the output file. In this example: ./test.mp3. If no output file is defined, everything is sent to the standard output ('stdout'). Usually, this is the xterm in which the program is executed. All mp3-codes are plotted on your screen instead of being stored in a file...

Please take care that your mixer-settings are set correctly. If not, nothing will be recorded! The correct recording-source should be selected and the appropiate volume-controls should be at their highest settings.

How to use 'sox | lame'

Type the following command in an xterm:
sox -t ossdsp -w -s -r 44100 -c 2 /dev/dsp -t raw - | lame -x -m s - ./test.mp3
Explanation of the supplied options:
sox The recording program: sound is recorded using the 'line-in' of the soundcard.
-t ossdsp The sound input originates from the OSS sound drivers. These drivers are included with the Linux kernel if the standard sound support is included.
-w Sample bit type is 16 bit long word (if you do not understand this: don't bother, it is not important).
-s The 'sample data' is 'signed linear' (if you do not understand this: don't bother, it is not important).
-r 44100 The sampling frequency is 44100 Hz (CD-quality).
-c 2 Record from 2 channels (stereo).
/dev/dsp This is the sound-sampling device: the Linux-device that delivers the recording data. All devices, ports, disks, and so on are accessed using the '/dev'-directory. The wave-stream is accessed in linux from '/dev/dsp'. As a matter of fact, it is even possible to get your soundfile using the following command: 'cat /dev/dsp >test.wav' (works only if the mixer-settings are correct!!!).
-t raw The output is 'raw' data (unchanged, straight from the soundcard).
- Use the standard output, symbolised by '-'
lame The encoder program: the recorded sound is piped to 'lame' for the mp3 conversion.
-x 'Force byte-swapping of input'. I do not know the precise function of this option, but without it the procedure simply does not work (before I found this out!).
-m s Encode as stereo mp3-file. Standard, 'lame' encodes it as 'joint-stereo': only the DIFFERENCES between both sound channels are encoded. As a result, a higher compression is obtained, but the mp3-quality is poorer.
- Input is 'stdin'. This is correct for the current example.
./test.mp3 The name of the final mp3-file.

The mixer

A correct mixer-setting is very important. The mixer determines the recording source. Standard, the KDE-mixer is set as follows:
The standard KDE mixer-settings
Figure 1: The standard KDE mixer-settings.

Figure 1 depicts that the microfone is the current recording-source. For our purpose, an external source is connected to the computer using the 'line-in' connector on the soundcard. The recording-source should be the 'line-in'. If the mouse-pointer is positioned on the line-in channel of the KDE-mixer, a right-click results in the menu that is depicted in Figure 2:

Selecting the correct recording-source
Figure 2: selecting the correct recording-source.

Off course, 'RecSource' is selected to set the 'line-in' of the soundcard as the recording-source. Use the same procedure to turn the microfone off as a recording-source: you do not want to have the additional noise!


If the 'line-in' is not selected as recording-source, no sound is obtained from the 'line-in' (even if you can hear the music yourself)!!! Do not forget to set the 'line-in' on 'unmute'. Figure 2 depicts this correctly: you can set the 'line-in' on mute, so currently it is 'unmuted'.

The KDE-mixer is set correctly
Figure 3: The KDE-mixer is set correctly.

Precautions to obtain good recordings

It is likely that you usually work in XWindows and that 'mpegrec' will be executed in XWindows. This is not a good idea for the following two reasons:

The remarks above indicate that recordings can be obtained best if no graphical environment is present. This means that the Bash-shell should so used (a text-based shell). This approach has some consequences:

These two problems can be solved quite easily. The first problem, sound support, is solved as follows:
'modprobe -k sb'

This command loads the module 'sb', which installs my soundcard (a SoundBlaster AWE 64). If you do not know which module should be installed, please look in your '/etc/conf.modules' (SuSE) or '/etc/modules.conf' (depending on your distribution). For me, the soundcard part is the following:

alias char-major-14 sb
pre-install sb /sbin/modprobe "-k" "adlib_card"
pre-install sb /sbin/modprobe "-k" "awe_wave"
options sb io=0x220 irq=5 dma=1 dma16=5 # mpu_io=0x330
options adlib_card io=0x388 # FM synthesizer
post-install sb /usr/bin/sfxload "/win98/Program Files/Creative/ctsnd/sfbank/synthgs.sbk"
Your soundcard is always defined in 'alias char-major-14' (if the standard OSS-drivers are used, which is usually the case). This line is closed with 'sb', so I know that the command 'modprobe -k sb' initializes the soundcard services.

The second problem is selecting the 'line-in' on your soundcard with a text-based mixer-program. I use the 'aumix' program. If in Bash the command 'aumix' is executed, the mixer starts in interactive mode. In this mode, settings can be checked and changed in a glance, see Figure 4.

The text-oriented mixer 'aumix'
Figure 4: The text-oriented mixer 'aumix'.

You can select the different recording sources with the arrow buttons. The '<' or '>' buttons selects the volume (left) or the balance (right). The '+' and '-' buttons increases and decreases the volume of the currently selected source. 'm' mutes or unmutes the current source. If the current source is selected as recording source, a red 'R' is displayed on the left of this volume source. If the source only can be used for playback, a green 'P' is displayed. Pressing 's' saves the current settings in '~/.aumixrc'. These settings are activated very easily on the next startup with the command: 'aumix -l'.

To automate the above procedure, I have written a small script: './mrec test.mp3'. This script contains the following lines:

if [ -z "$1" ]; then
	echo "Usage: mrec mp3-output-file"
	exit 0
modprobe -k sb
aumix -l
mpegrec -b 128 -o "$1"

Sound is initialized, 'line-in' is selected as recording-source, and 'mpegrec' starts immediately with recording.

Installing mpegrec

I made the 'mpegrec' software used in this article available for download. More information regarding 'mpegrec' can be found on Only the source-files and the compiled win32 versions can be found here. A page with more information does not exist, but some more information can be found here.

The 'lame'-files and installation tips can be found here.

Installation under Linux

Install lame

The 'lame' version used in this article is as well available for download. The 'lame' homepage is on

Linux installation

Installing sox

'Sox' is a program to record, convert, or edit sound. As an example, an echo can be added to a sound file. Read the 'sox' man-pages ('man sox') for more information. Conversions that 'sox' can do are: convert a wave-file to an .au-file (the standard SUN audio format) and vice-versa. This possibility is sometimes very handy! The version of 'sox' used for this article can be >found here< and the homepage is at

Linux installation

Other problems

I found two other problems:


This is a general SuSE problem. A heavy process is executed to update the man-pages and to remove old log-files. This process is very useful, but uses to much processor power for mp3-recording. I changed the name of 'cron' during recording, so it could not be found and executed. After recording, 'cron' is renamed back! Use the following command to change the name of 'cron':

mv /usr/sbin/cron /usr/sbin/cron.backup

It is likely that the location of 'cron' differs for other Linux distributions. You can find the location of 'cron' with the following command:

which cron

Please check whether 'cron' is needed for other tasks! On a desktop Linux machine, this should not be a problem.

Sound driver problem

This problem was solved by using the alternative Alsa sound drivers. Use the Alsa documentation (not included in the packed driver/utilities files) to obtain instructions on how to install these drivers. It is better not to use these drivers unless real problems do occur.