mpg123(1)

📦
version 1.33.7 · 2 versions
, in alpine 3.23, alpine 3.24, arch rolling, nixpkgs unstable, ubuntu stonking

💖 mpg123 — the girlies' guide to playing MPEG audio 💖

okay bestie, so this is the manual page for mpg123, and it's literally the tool for playing audio MPEG 1.0/2.0/2.5 streams (layers 1, 2 and 3). she's iconic, she's been around forever, and she's about to be your new fave. 🎀

✨ SYNOPSIS ✨

mpg123 [ options ] file-or-URL...

🌸 DESCRIPTION 🌸

so mpg123 reads one or more files (or standard input, if you pass -) or URLs, and then either plays them on the audio device (that's the default, bestie 💅) or outputs them to stdout. and every file/URL is assumed to be an MPEG audio bit stream. no cap.

🦋 OPERANDS 🦋

here's what you can actually hand her:

  • file(s) — the path name(s) of one or more input files. they must be valid MPEG-1.0/2.0/2.5 audio layer 1, 2 or 3 bit streams. if you specify a dash -, the MPEG data gets read from the standard input instead. also, any name starting with http:// or https:// is recognized as a URL (peep the next section), while a leading file:// is being stripped for normal local file access, for consistency (since mpg123 1.30.1). periodt.

💅 OPTIONS 💅

okay so mpg123 options may be either the traditional POSIX one letter options, or the GNU style long options. POSIX style options start with a single -, while GNU long options start with --. option arguments (if needed) follow separated by whitespace — and not =, so don't try that one. also, note that some options can be absent from your installation when they got disabled in the build process. lowkey tragic, but that's how it is. 🙈

🌈 INPUT OPTIONS 🌈

  • -k num, --skip num — skip the first num frames. by default the decoding starts at the first frame.
  • -n num, --frames num — decode only num frames. by default the complete stream is decoded.
  • --fuzzy — enable fuzzy seeks (guessing byte offsets or using approximate seek points from Xing TOC). without that, seeks need a first scan through the file before they can jump at positions. so this is your choose-your-fighter moment: sample-accurate operation with gapless features, or faster (fuzzy) seeking. pick your vibe. ✨
  • -y, --no-resync — do NOT try to resync and continue decoding if an error occurs in the input file. normally, mpg123 tries to keep the playback alive at all costs, including skipping invalid material and searching a new header when something goes wrong. with this switch you can make it bail out on data errors (and perhaps spare your ears a bad time). note that this switch has been renamed from --resync — the old name still works, but is not advertised or recommended to use (subject to removal in future). so like, use the new one, bestie. 💀
  • -F, --no-frankenstein — disable support for Frankenstein streams. normally, mpg123 stays true to the concept of MPEG audio being just a concatenation of MPEG frames. she'll continue decoding even if the type of MPEG frames varies wildly. with this switch, she will only decode the input as long as it does not change its character (from layer I to layer III, changing sampling rate, from mono to stereo), silently assuming end of stream on such occasion. the switch also stops decoding of compatible MPEG frames if there was an Info frame (Xing header, Lame tag) that contained a length of the track in MPEG frames. this comes a bit closer to the notion of a MP3 file as a defined collection of MPEG frames that belong together, but it gets rid of the flexibility that can be fun at times but mostly is hell for the programmer of the parser and decoder... 😭
  • --network backend — select network backend (helper program), choices are usually auto, wget, and curl. auto means to try the first available backend.
  • --resync-limit bytes — set the number of bytes to search for valid MPEG data once lost in stream; a value <0 means search the whole stream. if you know there are huge chunks of invalid data in your files... here is your hammer. 🔨 note: only since version 1.14 does this also increase the amount of junk skipped on beginning.
  • -u auth, --auth auth — HTTP authentication to use when receiving files via HTTP. the format used is user:password. mpg123 will clear this quickly, but it may still appear in sight of other users or even just in your shell history. 👀 you may seek alternative ways to specify that to your network backend.
  • --auth-file authfile — provide the authentication info via the given file instead of the command line directly. much more discreet, we love that for you. 🫶
  • --ignore-mime — ignore MIME types given by the HTTP server. if you know better and want mpg123 to decode something the server thinks is image/png, then just do it. main character energy.
  • --no-icy-meta — do not accept ICY meta data.
  • --streamdump filename — dump a copy of the input data (as read by libmpg123) to the given file. this enables you to store a web stream to disk while playing, or just create a concatenation of the local files you play for ... why not? 💖
  • --icy-interval bytes — this setting enables you to play a stream dump containing ICY metadata at the given interval in bytes (the value of the icy-metaint HTTP response header). without it, such a stream will play, but will cause regular decoding glitches with resync. so like, set it, bestie.
  • --no-seekbuffer — disable the default micro-buffering of non-seekable streams that gives the parser a safer footing.
  • -@ file, --list file — read filenames and/or URLs of MPEG audio streams from the specified file, in addition to the ones specified on the command line (if any). note that file can be either an ordinary file, a dash - to indicate that a list of filenames/URLs is to be read from the standard input, or an URL pointing to an appropriate list file. note: only one -@ option can be used (if more than one is specified, only the last one will be recognized). furthermore, for HTTP resources, the MIME type information will be used to re-open an actual MPEG stream as such instead of treating it as a playlist file. so you could just always use -@ for web resources without bothering if it is a playlist or already the resolved stream address. so convenient, honestly. ✨
  • -l n, --listentry n — of the playlist, play the specified entry only. n is the number of the entry starting at 1. a value of 0 is the default and means playing the whole list, a negative value means showing of the list of titles with their numbers...
  • --continue — enable playlist continuation mode. this changes frame skipping to apply only to the first track and also continues to play following tracks in the playlist after the selected one. also, the option to play a number of frames only applies to the whole playlist. basically, this tries to treat the playlist more like one big stream (like, an audio book). the current track number in list (1-based) and frame number (0-based) are printed at exit (useful if you interrupted playback and want to continue later). note that the continuation info is printed to standard output unless the switch for piping audio data to standard out is used. also, it really makes sense to work with actual playlist files instead of lists of file names as arguments, to keep track positions consistent. this interacts fine with --random since version 1.33, but not --shuffle, as that destroys the notion of a stable position in the playlist. 😭
  • --loop times — for playing track(s) a certain number of times in a loop, < 0 means infinite loop. forever and ever, bestie. 🔁
  • --keep-open — for remote control mode: keep the loaded file open after reaching the end.
  • --timeout seconds — timeout in (integer) seconds before declaring a stream dead (if <= 0, wait forever).
  • -z, --shuffle — shuffle play. randomly shuffles the order of files specified on the command line, or in the list file.
  • -Z, --random — continuous random play. keeps picking a random file from the command line or the play list. unlike shuffle play above, random play never ends, and plays individual songs more than once. she's obsessed with that one song fr. 🔥
  • -i, --index — index / scan through the track before playback. this fills the index table for seeking (if enabled in libmpg123) and may make the operating system cache the file contents for smoother operating on playback.
  • --index-size size — set the number of entries in the seek frame index table.
  • --preframes num — set the number of frames to be read as lead-in before a seeked-to position. this serves to fill the layer 3 bit reservoir, which is needed to faithfully reproduce a certain sample at a certain position. note that for layer 3, a minimum of 1 is enforced (because of frame overlap), and for layer 1 and 2, this is limited to 2 (no bit reservoir in that case, but engine spin-up anyway).

okay that's the input side of things, and there's more coming — don't go anywhere bestie, we're only getting started. 💜

  • -o module, --output module — picks which audio output module you're using. Bestie you can literally hand it a comma-separated list and mpg123 will take the first one that actually works. Also go look at -a, they're besties.
  • --list-modules — lists all the modules you have available. Window shopping, but for audio backends. 🛍️
  • --list-devices — lists the available output devices for whichever output module you've got going. Heads up though: if the chosen module has no device-listing functionality at all, you get an error printed and mpg123 bails with a non-zero exit code. No cap.
  • -a dev, --audiodevice dev — tells it which audio device to use. The default and the possible values both depend on which output is currently active, so it's giving "it depends" energy. For the JACK output specifically, you can pass a comma-separated list of ports to connect to — one for each channel. 🎧
  • -s, --stdout — the decoded audio samples get written to standard output instead of being played through your audio device. You must use this one if your audio hardware just isn't supported by mpg123. The output format defaults to raw (headerless) linear PCM audio data, 16 bit, stereo, host byte order — and yes, you can force mono or 8bit if that's your thing.
  • -O file, --outfile — writes raw output straight into a file, instead of doing the whole "redirect standard output to a file with the shell" dance. 🫶
  • -w file, --wav — writes the output as a WAV file. This makes the MPEG stream get decoded and saved as file, or to standard output if you use - as the filename. You can also reach for --au and --cdr for AU and CDR format respectively. Now, real talk: WAV/AU writing to non-seekable files, or to a redirected stdout, needs some thought. Since 1.16.0 the logic changed so the header gets written along with the first actual data — which avoids spurious WAV headers sitting in a pipe, for example. The result of decoding literally nothing to WAV/AU is a file that consists of just the header when it's seekable, and genuinely nothing at all when it isn't (not even a header). Correctly writing data with prophetic headers to stdout is no easy business. 😭
  • --au file — doesn't play the MPEG file, it writes it to file in SUN audio format instead. If you use - as the filename, the AU file goes to stdout. See the paragraph about WAV writing for the header drama with non-seekable streams.
  • --cdr file — doesn't play the MPEG file, writes it to file as a CDR file instead. If - is the filename, the CDR file gets written to stdout.
  • --reopen — forces a reopen of the audiodevice after every single track. Every. One. 💅
  • --cpu decoder-type — selects a certain decoder, one that's optimized for a specific CPU — think i586 or MMX. The list of available decoders can vary depending on the build and on what your CPU actually supports. This option only exists when the build actually includes several optimized decoders.
  • --test-cpu — tests your CPU and prints out a list of possible choices for --cpu. Vibe check your silicon. 🧠
  • --list-cpu — lists all available decoder choices, regardless of whether your CPU supports them.
  • -g gain, --gain gain — [DEPRECATED] sets the audio hardware output gain (default: don't change it). The unit of the gain value is hardware and output module dependent. This parameter is only hanging around for backwards compatibility and may be removed in the future without prior notice. Use the audio player for playing and a mixer app for mixing, UNIX style! periodt.
  • -f factor, --scale factor — changes the scale factor (default: 32768).
  • --rva-mix, --rva-radio — enables RVA (relative volume adjustment) using the values stored for ReplayGain radio mode / mix mode, where all tracks are roughly equal loudness. The first valid information it finds gets used — either in ID3V2 Tags (the Comment named RVA or the RVA2 frame) or the ReplayGain header in the Lame/Info Tag.
  • --rva-album, --rva-audiophile — enables RVA (relative volume adjustment) using the values stored for ReplayGain audiophile mode / album mode, which usually has the effect of adjusting album loudness while keeping the relative loudness inside the album intact. The first valid information found gets used — either in ID3V2 Tags (the Comment named RVA_ALBUM or the RVA2 frame) or the ReplayGain header in the Lame/Info Tag.
  • -0, --single0; -1, --single1 — decodes only channel 0 (left) or channel 1 (right), respectively. These are available for stereo MPEG streams only.
  • -m, --mono, --mix, --singlemix — mixes both channels / decodes mono. It takes less CPU time than full stereo decoding. Lowkey efficient. ✨
  • --stereo — forces stereo output.
  • -r rate, --rate rate — sets the sample rate (default: automatic). You might want to change this if you need a constant bitrate that's independent of the mpeg stream rate — mpg123 automagically converts the rate for you. If you do that, you should combine it with --stereo or --mono. 🌈
  • --resample method — sets the resampling method to use if you're forcing an output rate. The choices (case-insensitive) are NtoM, dirty, and fine. The fine resampler is the default — it employs libsyn123's low-latency, fairly efficient resampler to postprocess the output from libmpg123, instead of the fast but very crude NtoM decoder (the drop sample method) that mpg123 has been offering since decades ago. If you are really low on CPU time, choose NtoM, because the resampler usually needs more time than the MPEG decoder itself. The mpg123 program is smart enough to combine the 2to1 or 4to1 downsampling modes with the postprocessing for extreme downsampling. 💀
  • -2, --2to1; -4, --4to1 — performs a downsampling of ratio 2:1 (22 kHz from 44.1 kHz) or 4:1 (11 kHz) on the output stream, respectively. Saves some CPU cycles, but of course it throws away the high frequencies, since the decoder doesn't bother producing them in the first place.
  • --pitch value — sets a pitch change (speedup/down — 0 is neutral; 0.05 is a 5% speedup). When you're not enforcing an output rate, this changes the output sampling rate, so it only works within the range your audio system/hardware supports. When you combine it with a fixed output rate, it modifies a software resampling ratio instead.
  • --8bit — forces 8bit output.
  • --float — forces f32 encoding.
  • -e enc, --encoding enc — chooses the output sample encoding. Possible values look like f32 (32-bit floating point), s32 (32-bit signed integer), u32 (32-bit unsigned integer) and the variants with different numbers of bits (s24, u24, s16, u16, s8, u8), plus special variants like ulaw and alaw 8-bit. Peep the output of mpg123's longhelp for the encodings that are actually available.
  • -d n, --doublespeed n — only plays every n'th frame. This makes the MPEG stream play n times faster, which you can use for special effects. Can also be combined with the --halfspeed option to play 3 out of 4 frames, etc. Don't expect great sound quality when using this option.
  • -h n, --halfspeed n — plays each frame n times. This makes the MPEG stream play at 1/n'th speed (n times slower), which you can use for special effects. Can also be combined with the --doublespeed option to double every third frame or things like that. Don't expect great sound quality when using this option.
  • -E file, --equalizer — enables equalization, taken from file. The file needs to contain 32 lines of data, and additional comment lines may be prefixed with #. Each data line consists of two floating-point entries, separated by whitespace — they specify the multipliers for the left and right channel of a certain frequency band, respectively. The first line corresponds to the lowest frequency band, the 32nd to the highest. Note that you can control the equalizer interactively with the generic control interface. Also note that these are the 32 bands of the MPEG codec, not spaced like you would see for a usual graphic equalizer. The upside is that there's zero computational cost in addition to decoding. The downside is that you roughly have bass in band 0, (upper) mids in band 1, and treble in all the others. 😅
  • --gapless — enables code that cuts (junk) samples at the beginning and end of tracks, which enables gapless transitions between MPEG files when encoder padding and codec delays would otherwise prevent it. This is enabled per default beginning with mpg123 version 1.0.0. 🎀
  • --no-gapless — disables the gapless code. That gives you MP3 decodings that include encoder delay and padding plus mpg123's decoder delay.
  • --no-infoframe — do not parse the Xing/Lame/VBR/Info frame; decode it instead just like a stupid old MP3 hardware player. This implies disabling of gapless playback, since the necessary information lives in that very metadata frame.
  • -D n, --delay n — inserts a delay of n seconds before each track.
  • -o h, --headphones — directs audio output to the headphone connector (some hardware only; AIX, HP, SUN).
  • -o s, --speaker — directs audio output to the speaker (some hardware only; AIX, HP, SUN).
  • -o l, --lineout — directs audio output to the line-out connector (some hardware only; AIX, HP, SUN).
  • -b size, --buffer size — uses an audio output buffer of size Kbytes. This is useful for bypassing short periods of heavy system activity, which would normally cause the audio output to get interrupted. You should specify a buffer size of at least 1024 (i.e. 1 Mb, which equals about 6 seconds of audio data) or more; anything less than about 300 doesn't make much sense. The default is 0, which turns buffering off. 🦋
  • --preload fraction — waits for the buffer to be filled to fraction before starting playback (fraction is between 0 and 1). You can tune this prebuffering to either get faster sound to your ears, or safer uninterrupted web radio. Default is 0.2 (wait for 20 % of buffer to be full, changed from 1 in version 1.23).
  • --devbuffer seconds — sets the device buffer in seconds; <= 0 means the default value. This is the small buffer between the application and the audio backend, possibly directly related to hardware buffers.
  • --smooth — keeps the buffer over track boundaries — meaning, it does not empty the buffer between tracks, for possibly some added smoothness. 🌸

✨ MISC OPTIONS ✨

  • -t, --test — test mode, bestie! 🧪 The audio stream gets decoded and everything, but like… no output happens. It's giving silent disco.
  • -c, --check — check for filter range violations (that's clipping, if you were wondering 👀) and report them for each frame if any show up.
  • -v, --verbose — crank the verbosity up a notch. Like, it'll show you the frame numbers while decoding. Very "let me tell you everything" energy.
  • -q, --quiet — the opposite vibe. Suppress diagnostic messages. Shhh. 🤫
  • -C, --control — enable terminal control keys. This turns itself on automatically if a terminal is detected, no cap. By default you use s or the space bar to stop/restart (pause, unpause) playback, f to jump forward to the next song, b to jump back to the beginning of the song, , to rewind, . to fast forward, and q to quit. Type h for the full list of available controls, iykyk. The A-B loop feature with key o changes the preset loop interval to the interval between two presses of o, and the third press (or p) ends the looped playback. After that, the key p will use the updated loop interval. 💅
  • --no-control — disable terminal control even if a terminal is detected. Boundaries, periodt.
  • --title — in an xterm, rxvt, screen, iris-ansi (compatible — the TERM environment variable is examined, bestie), it changes the window's title to the name of the song currently playing. Main character energy for your terminal window 🎀
  • --pauseloop seconds — set the length of the loop interval in terminal control fixed looping mode, away from the default of 0.5 seconds, as a floating point number. This value can be overwritten at runtime using the A-B loop feature. Flexible queen ✨
  • --name name — set the name of this instance, possibly used in various places. This sets the client name for JACK output.
  • --long-tag — display ID3 tag info always in long format, with one line per item (artist, title, …). No cramming, we're doing full sentences here.
  • --utf8 — regardless of environment, print metadata in UTF-8. Otherwise, when you're not using a UTF-8 locale, you'll get the ASCII stripdown. Sad! 😭
  • -R, --remote — activate the generic control interface. mpg123 will then read and execute commands from stdin. Basic usage is load <filename> to play some file and the obvious pause, command. jump <frame> will jump/seek to a given point (MPEG frame number). Issue help to get a full list of commands and syntax. So handy 🫶
  • --remote-err — print responses for generic control mode to standard error, not standard out. This is automatically triggered when using -s.
  • --fifo path — create a fifo / named pipe on the given path and use that for reading commands instead of standard input. So very unix of you 💜
  • --aggressive — tries to get higher priority. Go get it, girl 🔥
  • -T, --realtime — tries to gain realtime priority. This option usually requires root privileges to have any effect. (So like… ask permission first 🙈)
  • -?, --help — shows short usage instructions.
  • --longhelp — shows long usage instructions. The director's cut 🎬
  • --version — print the version string.
  • --libversion — print version information on the mpg123 libraries being used (libmpg123, libout123, libsyn123). The whole squad 🌈

🌐 HTTP SUPPORT 🌐

So besides reading MPEG audio streams from ordinary files and from standard input, mpg123 also supports retrieving MPEG audio streams or playlists via the HTTP protocol — that's the one used on the World Wide Web (WWW), bestie. Such files are specified using a so-called URL, which starts with http:// or https://. When a file with that prefix is encountered, mpg123 since 1.30.0 will by default call an external helper program (either wget(1) or curl(1), see the --network option) to retrieve the resource. You can configure access via a proxy server using the standard environment variables those programs support. The --proxy option that mpg123 before 1.30.0 used for its internal network code is gone in the default build now and will probably disappear for good with 1.31.1. Rip to that one 💀

Note that, in order to play MPEG audio files from a WWW server, it is necessary that the connection to that server is fast enough. For example, a 128 kbit/s MPEG file requires the network connection to be at least 128 kbit/s (16 kbyte/s) plus protocol overhead. If you suffer from short network outages, you should try the -b option (buffer) to bypass such outages. If your network connection is generally not fast enough to retrieve MPEG audio files in realtime, you can first download the files to your local harddisk (e.g. using wget(1)) and then play them from there.

Streams with embedded ICY metadata are supported — the interval being communicated via HTTP headers or --icy-interval. Slay 🎧

🛑 INTERRUPT 🛑

When in terminal control mode, you can quit via pressing the q key, while any time you can abort mpg123 by pressing Ctrl-C. If not in terminal control mode, this will skip to the next file (if any). If you want to abort playing immediately in that case, press Ctrl-C twice in short succession (within about one second). Double tap, fr fr ✌️

Note that the result of quitting mpg123 pressing Ctrl-C might not be audible immediately, due to audio data buffering in the audio device. This delay is system dependent, but it is usually not more than one or two seconds. She's just finishing her sentence, okay? 😌

📊 PLAYBACK STATUS LINE 📊

In verbose mode, mpg123 updates a line with various information centering around the current playback position. On any decent terminal, the line also works as a progress bar in the current file by reversing video for a fraction of the line according to the current position. Aesthetic and functional 💅

An example for a full line is this:

> 0291+0955  00:01.68+00:28.22 [00:05.30] mix 100=085 192 kb/s  576
B acc   18 clip p+0.014

The information consists of, in order:

  • > — single-character playback state (> for playing, = for pausing/looping, _ for stopped)
  • 0291+0955 — current frame offset and number of remaining frames after the plus sign
  • 00:01.68+00:28.22 — current position from and remaining time in human terms (hours, minutes, seconds)
  • [00:05.30] — fill of the output buffer in terms of playback time, if the buffer is enabled
  • mix — selected RVA mode (possible values: mix, alb (album), and --- (neutral, off))
  • 100=085 — set volume and the RVA-modified effective volume after the equal sign
  • 192 kb/s — current bitrate
  • 576 B — size of current frame in bytes
  • acc — if positions are accurate, possible values are acc for accurate positions or fuz for fuzzy (with guessed byte offsets using mean frame size)
  • 18 clip — amount of clipped samples, non-zero only if decoder reports that (generic does, some optimized ones not)
  • p+0.014 — pitch change (increased/decreased playback sampling rate on user request)

📝 NOTES 📝

MPEG audio decoding requires a good deal of CPU performance, especially layer-3. To decode it in realtime, you should have at least an i486DX4, Pentium, Alpha, SuperSparc or equivalent processor. You can also use the -m option to decode mono only, which reduces the CPU load somewhat for layer-3 streams. See also the -2 and -4 options. Choose your fighter ⚔️

If everything else fails, have mpg123 decode to a file and then use an appropriate utility to play that file with less CPU load. Most probably you can configure mpg123 to produce a format suitable for your audio device (see above about encodings and sampling rates).

If your system is generally fast enough to decode in realtime, but there are sometimes periods of heavy system load (such as cronjobs, users logging in remotely, starting of big programs etc.) causing the audio output to be interrupted, then you should use the -b option to use a buffer of reasonable size (at least 1000 Kbytes). Buffer her up, she's fragile 💖

🚪 EXIT CODE 🚪

Up to version 1.25.x, mpg123 always returned exit code 0 also for complete junk on the input side. Fatal errors were only considered for output. With version 1.26.0, this changed to the behaviour described below. Glow-up era ✨

When not using the remote control interface (which returns input errors as text messages), the process exit code is zero (success) only if all tracks in a playlist had at least one frame parsed — even if it did not decode cleanly, or are empty, MPEG-wise (perhaps only metadata, or really an empty file). When you decode nothing, nothing is the result and that is fine. When a track later aborts because of parser errors or breakdown of the network communication, this is treated as end of a track, but does not make the process as such fail. One really bad (or non-existing) stream in the playlist results in a non-zero error code, consistent with other UNIX tools.

An error in audio output results in the process ending with a non-zero exit code immediately, regardless of how much data has been successfully played before. The forgiveness is only on the input side. Output? She holds a grudge. 😤

🐛 BUGS 🐛

Mostly MPEG-1 layer 2 and 3 are tested in real life. Please report any issues and provide test files to help fixing them. We love a helpful bestie 🫶

No CRC error checking is performed. But the decoder is built and tested to behave nicely with damaged streams. Mostly, damaged frames will just be silent. Resilient queen 👑

Some platforms lack audio hardware support; you may be able to use the -s switch to feed the decoded data to a program that can play it on your audio device.

👑 AUTHORS 👑

Maintainer: Thomas Orgis <maintainer@mpg123.org>, <thomas@orgis.org>

Original Creator: Michael Hipp

Uses code or ideas from various people, see the AUTHORS file accompanying the source code. Respect to the whole group chat 💜

⚖️ LICENSE ⚖️

mpg123 is licensed under the GNU Lesser/Library General Public License, LGPL, version 2.1. Free as in freedom, bestie 🌈

🔗 WEBSITE 🔗

http://www.mpg123.org
http://sourceforge.net/projects/mpg123

And that's the whole page, babes 💖 This one's from the mpg123(1) manual, dated 11 Jul 2022. Go forth and decode, you absolute icon — no cap, you understood all of it. Periodt. 🎀✨

📦 where this page comes from