timidity - Man Page

MIDI-to-WAVE converter and player

Examples (TL;DR)


timidity [options] filename [...]


TiMidity++ is a converter that converts some of MIDI files (supported formats: Standard MIDI files (*.mid), Recomposer files (*.rcp, *.r36, *.g18, *.g36) and Module files (*.mod)) into formatted audio files (e.g. RIFF WAVE).  TiMidity++ uses Gravis Ultrasound-compatible patch files or Soundfonts (*.sfx, *.sf2) to generate digital audio data from MIDI files.  The digital audio data generated by TiMidity++ can be stored in a file for processing, or played in real time through an audio device.
In real time playing, TiMidity++ can show the lyrics contained in KAR or WRD files.


You can use the following expressions as the filename argument:


Read a MIDI file from standard input.


Read a MIDI file from the specified path on a filesystem.


Read and play all MIDI files in the specified directory.  For example,

% timidity some/where/

plays all files in the directory some/where/.

Archive File

Extract and play the file(s) in the archive.  If you want to specify a certain MIDI file in the archive, append #<MIDI-filename> to the archive name.  The path after `#' allows the use of the wildcard expressions (case insensitive).
You can use escape sequence \xHH, where `HH' is a ASCII number in hexadecimal integer.

For example:

% timidity file.zip#file.mid

Plays file.mid in file.zip

% timidity file.lzh#*.mid

Plays any files that match the wildcard expression *.mid in file.lzh

% timidity file.tgz#*

This expression is the same as file.tgz

Since these mechanism are contained in TiMidity++ itself, you can use this syntax even in the MS Windows environment.

TiMidity++ can handle the following archive formats:

tar (*.tar)
tar+gzip (*.tar.gz, *.tgz)
zip (*.zip)
lzh (*.lzh, *.lha)

(lh0, lh1, lh2, lh3, lh4, lh5, lh6, lz4, lzs and lz5 are available)

Other archives can be expanded if expander command is specified on the compile phase.  TiMidity++ will pipe that command.


Play the MIDI file in the specified article on the news server.  If a newsgroup is specified TiMidity++ plays all MIDI files found in any article posted to that newsgroup.
TiMidity++ parses MIME Multi-part messages in case of news://* scheme, extracts MIDI file that was post to that group, or in case of normal file, you can explicitly handle MIME documents by naming that file with *.mime extension, or prefix that file with mime:

The following MIME-types are allowed:

uu-encoded file

is required

base64 encoded

Content-Transfer-Encoding: base64
is required


Content-Transfer-Encoding: quoted-string
is required

Mac BinHex format

only HQX format is available


Play the file specified in the URL.

For example:

% timidity http://www.goice.co.jp/member/mo/dist/midi/impromptu.mid

plays the specified MIDI-file directly from the network.

If these expression are used in the *.cfg files, you can even use patch-files (and others) from remote machines.

Input File

TiMidity++ can handle the following file formats:

.mid, .rmi (Format 0, 1, 2)

Standard MIDI File

.rcp, .r36, .g18, .g36 (Recomposer formats)

Recomposer format which is product for COME ON MUSIC co.

.mfi (MFi Version 3 - Melody Format for i-Mode)

i-Mode is Japanese local mobile phone

.kar (Karaoke format)

Displays the lyrics as a Lyric Meta Event message.

.mod, mod.* (Module file)

.wrd (WRD format)


The following command line options are accepted by TiMidity++:

-A [n][,m](a)

Multiplies the master volume by n%.  Default value is 70%. Higher amplification makes louder sounds.  You can specify the drum power, ratio of drum volume from the other channels.  The allowed values of amplification range from 0 (no sound) to 800.
Optionally to put `a' character along with -a option, or to use --volume-compensation, instructs TiMidity++ to regularize the volume.  You can easily gain dynamic range.

For example:


volume 90%, drum power 100%, compensation is off


volume 70%, drum power 120%, compensation is off


volume 90%, drum power 120%, compensation is off


volume 70%, drum power 100%, compensation is on


volume 90%, drum power 100%, compensation is on


volume 70%, drum power 120%, compensation is on


volume 90%, drum power 120%, compensation is on

-a, --[no-]anti-alias

Turns on anti-aliasing.  Samples are run through a lowpass filter before playing, which reduces aliasing noise at low resampling frequencies.

-B n,m, --buffer-fragments=n,m

For the Linux/FreeBSD/OSS/ALSA/Windows sound driver, selects the number of buffer fragments in interactive mode.  Increasing the number of fragments may reduce choppiness when many processes are running. It will make TiMidity++ seem to respond sluggishly to fast forward, rewind, and volume controls, and it will throw the status display off sync.  Specify a fragments number of 0 to use the maximum number of fragments available.

-C n, --control-ratio=n

Sets the ratio of sampling and control frequencies.  This determines how often envelopes are recalculated -- small ratios yield better quality but use more CPU time.

-c file, --config-file=file

Reads an extra configuration file.

-D n, --drum-channel=n

Marks channel as a drum channel.  If channel is negative, channel -n is marked as an instrumental channel.  If n is 0, all channels are marked as instrumental.

-d dir, --interface-path=dir

Specifies the directory containing installed dynamic-link interface modules.

-E mode, --ext=mode

Set TiMidity++ extend modes.  The following modes are available (capitalized switch means disable this feature):

w/W, --[no-]mod-wheel

Enable/disable modulation controlling.

p/P, --[no-]portamento

Enable/disable portamento controlling.

v/V, --[no-]vibrato

Enable/disable NRPM vibration.

s/S, --[no-]ch-pressure

Enable/disable channel pressure controlling.

e/E, --[no-]mod-envelope

Enable/disable modulation envelope controlling.

t/T, --[no-]trace-text-meta

Enable/disable tracing all Text Meta Events.

o/O, --[no-]overlap-voice

Accept/reject pronouncing multiple same notes.

z/Z, --[no-]temper-control

Enable/disable MIDI Tuning Standard temperament controlling.

mHH, --default-mid=HH

Sets the manufacturer ID to HH (where HH are two hex-digits).
HH values of GS/gs, XG/xg or GM/gm are understood as 41, 43 and 7e respectively.

MHH, --system-mid=HH

Sets the system manufacturer ID to HH (where HH are two hex-digits).
In this option, the manufacture ID is set unchangeable. Manufacture ID from the input file would be ignored.

bn, --default-bank=n

Use tone bank n as the default.

Bn, --force-bank=n

Sets the bank number of all channels to n.

in[/m], --default-program=n[/m]

Use the program number as the default instrument.  Any Program Change events in MIDI files will override this option.
If n is followed by /m the default program number of the channel m is specified by n.

In[/m], --force-program=n[/m]

Similar to -Ei but this ignores all program changes.

F args

For effects.  See below.  In args option, you can specify following effect options:

delay=(d|l|r|b)[,msec], --delay=(d|l|r|b)[,msec]

Sets delay type.

d, 0

Disabled delay effect.

l, 1

Left delay.

r, 2

Right delay.

b, 3

Swap left & right.

Optional msec is the delay time.

chorus=(d|n|s)[,level], --chorus=(d|n|s)[,level]
d, 0

Disable this effect.

n, 1

Enable MIDI chorus effect control.

s, 2

Surround sound, chorus detuned to a lesser degree (default).

The optional parameter level specifies the chorus level 0 to 127.

reverb=(d|n|g)[,level], --reverb=(d|n|g)[,level]
d, 0

Disable MIDI reverb effect control.

n, 1

Enable MIDI reverb effect control.  This effect is only available in stereo.

g, 2

Global reverb effect.

f, 3

Freeverb MIDI reverb effect control (default).

G, 4

Global Freeverb effect.

The optional parameter level specifies the reverb level 0 to 127.
For Freeverb effect, additional parameters scaleroom, offsetroom and predelay can be specified. scaleroom and offsetroom specify variables for the room size calculated as
roomsize = C * scaleroom + offsetroom
where C is the reverberation character. predelay specifies the factor for pre-delay time of reverberation in percent. Any of the optional parameters can be omitted. Default values are 0.28, 0.7 and 100 respectively.

vlpf=(d|c|m), --voice-lpf=(d|c|m)
d, 0

Disable LPF effect.

c, 1

Chamberlin resonant LPF (12dB/oct) (default).

m, 2

Moog resonant low-pass VCF (24dB/oct)

ns=n, --noise-shaping=n

Enable the n th degree noiseshaping filter.  The distortion at decay stage is improved, but the noise on human auditory feeling increases because it shifts to a high frequency.  In case of 8-bit linear encoding, valid values of n are in the interval from 0 (min) to 4 (max).  Default value is 4.  In case of 16-bit linear encoding, valid values of n are in the interval from 0 to 4.  According to the value, it works as following. Default value is 4.


No noise shaping.


Traditional noise shaping.


Overdrive-like soft-clipping + new noise shaping.


Tube-amplifier-like soft-clipping + new noise shaping.


New noise shaping.

resamp=(d|l|c|L|n|g), --resample=(d|l|c|L|n|g)
d, 0

No interpolation.

l, 1

Linear interpolation.

c, 2

Cubic spline interpolation.

L, 3

Lagrange method.

n, 4

Newton polynomial interpolation.

g, 5

Modified Gauss effect (default).

This option affects the behavior of -N option.

-e,  --evil

Make TiMidity++ evil.  For the Win32 version, this increases the task priority by one.  It can give better playback when you switch tasks at the expense of slowing all other tasks down.

-F, --[no-]fast-panning

Turns on fast panning to accommodate MIDI pieces that expect panning adjustments to affect notes that are already playing.  Some files that don't expect this have a habit of flipping balance rapidly between left and right, which can cause severe popping when the -F flag is used.
In the current version of TiMidity++ this option is a toggle.

-f, --[no-]fast-decay

Toggles fast envelopes.  This option makes TiMidity++ faster but the release time of the notes are shortened.

-G begin1-end1[,begin2-end2,...](m)

Play just sub-segment(s) specified by comma separated time segments.
Each time segment defined by dash separated time values of: begin-end - defaulted to 0-infinity. Playing from begin to end.

Time format: [<minutes>:]<seconds>[.<milliseconds>]

To put `m' character along with the option treats it as measure and beat instead of seconds.

Time format: <measure>[.<beat>] (measure and beat are one-origin)

For example:


Skip 12 seconds and stop at 1 minute and half a second.


Play the initial 5 seconds, then skip to 2 minutes from start and play till end.

-G11.4-20m -G4:00-

Play from the 4th beat in the 11th measure to the 20th measure, then skip to 4 minutes from start and play till end.

-g sec, --spectrogram=sec

Open the Sound-Spectrogram window.  This option is activated if the system has support for the X Window System.

-H n, --force-keysig=n

Specify the number of key signature.  MIDI file is performed with transposition to the key of the number of sharp (when n is positive) or flat (when n is negative).  Valid values of n are in the interval from -7 to 7.  For example, if n is 1, MIDI file with any tonality will be performed transposing to 1 flat (i.e., F major or D minor).  Therefore, it is convenient to use for the accompaniment when practicing a musical instrument by the tonality which is easy to perform.

-h,  --help

Show the help message.

-i mode, --interface=mode

Selects the user interfaces from the compiled-in alternatives. mode must be begun with one of the supported interface identifiers.  Run TiMidity++ with the -h option to see a list.
For ALSA sequencer interface, optionally to use --realtime-priority, set the realtime priority by n, and to use --sequencer-ports, set the number of opened sequencer ports.  Default value is 4.
The following identifiers may be available:


dumb interface


ncurses interface


S-Lang interface


X Athena Widget interface


Tcl/Tk interface


Motif interface


vt100 interface


Emacs interface
(use “M-x timidity” in Emacs)


skin interface
Environment variable TIMIDITY_SKIN must be set to the path of the skin data (compressed data are also supported).


GTK+ interface


Launch TiMidity++ as MIDI server.


Launch TiMidity++ as ALSA sequencer client.


Windows synthesizer interface


Windows GUI interface


PortMIDI synthesizer interface


UMP interface

Interface options

Option characters may be added immediately after the interface identifier.  The following options are recognized:

v, --verbose=n

Increases verbosity level.  This option is cumulative.

q, --quiet=n

Decreases verbosity level.  This option is cumulative.

t, --[no-]trace

Toggles trace mode.  In trace mode, TiMidity++ attempts to display its current state in real time.  For the Linux sound driver, this is accomplished through the use of short DMA buffer fragments, which can be tuned via the -B option.

l, --[no-]loop

Loop playing (some interfaces ignore this option)

r, --[no-]random

Randomize file list arguments before playing

s, --[no-]sort

Sort file list arguments before playing

D, --[no-]background

Daemonize TiMidity++ in background (for alsaseq only)

-j, --[no-]realtime-load

Enable the loading of patch files during play.

-K n, --adjust-key=n

Adjusts key (i.e., transposes the song) by n half tones.  Ranges from -24 to 24.

-k msec, --voice-queue=msec

Specify audio queue time limit to reduce voices.  If the remaining audio buffer is less than msec milliseconds, TiMidity++ tries to kill some voices.  This feature makes it possible to play complicated MIDI files on slow CPUs.  Setting msec to zero tells TiMidity++ to never remove any voices.

-L path, --patch-path=path

Adds path to the library path.  Patch, configuration, and MIDI files are searched along this path.  Directories added last will be searched first.  Note that the current directory is always searched first before the library path.

-M name, --pcm-file=name

TiMidity++ can play a PCM file instead of a MIDI file.  If “auto” is specified, TiMidity++ tries to open foo.mid.wav or foo.mid.aiff when playing foo.mid.  If “none” is specified, this feature is disabled.  Otherwise just plays name.

-m msec, --decay-time=msec

Modify envelope volume decay time.  msec is the minimum number of milliseconds to sustain a sustained note.


Disable sustain ramping, causes constant volume sustains (default).


Effectively behaves as if all sustains are ignored, volume ramping is the same as normal stage 3.


A note at full volume will decay for 3 seconds once it begins to be sustained (assuming the regular stage 3 rate would not cause it to decay even longer).  Softer notes will of course die sooner.

-N n, --interpolation=n

Sets interpolation parameter.  This option depends on the -EFresamp option's value.

cspline, lagrange

Toggles 4-point linear interpolation (default is on).


n point interpolation using Newton polynomials.  n must be an odd number from 1 to 57.


n+1 point modified Gauss interpolation.  Ranges 0 (disable) to 34 (max), default to 25.

In either way, linear interpolation is used if audio queue < 99%.

-O mode, --output-mode=mode

Selects the output mode from the compiled-in alternatives. mode must begin with one of the supported output mode identifiers.  Run TiMidity++ with the -h option to see the list.
Special in Ogg FLAC output mode, verifying generated data (will be a bit slower), the size of header padding (default is 4096), the compression level (0 to 8) (default is 5), and enabling OggFLAC stream can be specified by --flac-verify, --flac-padding, --flac-complevel and --oggflac options respectively.
Special in Ogg Speex output mode, the compression quality (0 to 10) (default is 8), Enabling VBR output, enabling ABR output and setting the ratio to n, enabling VAD (voice activity detection), enabling DTX (discontinuous transmission), the encoding complexity (0 to 10) (default is 3), and frames in a single Ogg packet (0 to 10) (default is 1) can be specified by --speex-quality, --speex-vbr, --speex-abr, --speex-vad, --speex-dtx, --speex-complexity and --speex-nframes options respectively.
The following identifiers are available in all versions:


Outputs via OSS dsp device (default)


Output to ALSA


Generate raw waveform data.  All format options are supported.  Common formats include:




16-bit signed linear PCM


8-bit unsigned linear PCM


Generate Sun Audio (au) data


Generate AIFF data


Generate RIFF WAVE format output.  If output is directed to a non-seekable file, or if TiMidity++ is interrupted before closing the file, the file header will contain 0xffffffff in the RIFF and data block length fields.  The popular sound conversion utility sox is able to read such malformed files, so you can pipe data directly to sox for on-the-fly conversion to other formats.


List MIDI events


MOD -> MIDI conversion












Ogg Vorbis




Ogg Speex



Format options

Option characters may be added immediately after the mode identifier to change the output format.  The following options are recognized:

S, --output-stereo


M, --output-mono


s, --output-signed

Signed output

u, --output-unsigned

Unsigned output

1, --output-16bit

16-bit sample width

2, --output-24bit

24-bit sample width

8, --output-8bit

8-bit sample width

l, --output-linear

Linear encoding

U, --output-ulaw

u-Law (8-bit) encoding

A, --output-alaw

A-Law encoding

x, --[no-]output-swab

Byte-swapped output

Note that some options have no effect on some modes.  For example, you cannot generate a byte-swapped RIFF WAVE file, or force uLaw output on a Linux PCM device.

-o file, --output-file=file

Place output on file, which may be a file, device, or HP-UX audio server, depending on the output mode selected with the -O option.  The special filename `-' causes output to be placed on stdout.

-P file, --patch-file=file

Use patch file for all programs.

-p [n](a)

Sets polyphony (maximum number of simultaneous voices) to n.
Optionally to put `a' character along with -p option, or to use --polyphony-reduction, instructs TiMidity++ to enable automatic polyphony reduction algorithm.

-Q n[,...](t)

Cause channel n to be quiet.  n can carry out package specification by `,'.  If n is 0, all channels are turned off.  Continuously, specifying -n, channel n is turned back on.
On the other hand, to put `t' character after -Q option or to use --temper-mute describes temperament mute.  This mutes channels of specific temperament type n.  For preset temperament, n can range 0 to 3.  For user-defined temperament, n can range 4 to 7.

-q sec/n, --audio-buffer=sec/n

Specify audio buffer in seconds.  sec maximum size of buffer, n percentage filled at the beginning (default is 5.0/100) (size of 100% equals the whole device buffer size).

-R msec

Enables Pseudo Reverb Mode.  It sets every instrument's release to msec ms.  If msec is 0, msec is set to 800 (default).

-S n, --cache-size=n

Sets the re-sample cache size to n bytes.  If n equals 0 any sample caches are disabled.  The default value of n is 2097152 (2MB).

-s freq, --sampling-freq=freq

Sets the resampling frequency (Hz or kHz).  Not all sound devices are capable of all frequencies -- an approximate frequency may be selected, depending on the implementation.

-T n, --adjust-tempo=n

Adjust tempo to n%; 120 play MOD files with an NTSC Amiga's timing.

-t code, --output-charset=code

Sets output coding of Japanese text.  Possible values of code are:


determined by the LANG environment variable.


Translates non-ASCII code to period.


No conversion.


Convert from windows-1251 to koi8-r.


Outputs EUC (Japan) coding.


Outputs JIS coding.


Outputs SJIS coding.

-U, --[no-]unload-instruments

Unload all instruments from memory between MIDI files.  This can reduce memory requirements when playing many files in succession.

-V power, --volume-curve=power

Set the power of volume curve.  The total amplification becomes volume^power.  0 (default) uses the regular tables.  Any non-zero value causes all midi to use the new user defined velocity/volume/expression curve (linear: 1, ideal: ~1.661, GS: ~2).

-v,  --version

Show the version string

-W mode, --wrd=mode

Play WRD file.

Allowed values of mode are:


X Window System mode


Windows console mode


TTY mode


Dumb mode (outputs WRD events directory)


do not trace WRD


Sets WRD options:


Sets the WRD options.  an is the name of option and bn is the value.


Emulates timing (@WAIT, @WMODE) bugs of the original MIMPI player. The emulation levels are:


do not emulate any bugs of MIMPI


only emulate some bugs (default)


emulate all known bugs


Use file as WRD file only no file matching *.wrd is found.


Uses file as WRD file.

WRD mode must also use trace mode (option -i?t) or the timing of the WRD events will be terrible.

-w mode, --rcpcv-dll=mode

Extended mode for MS Windows.  The following options are available:

-w r

Use rcpcv.dll to play RCP/R36 files.

-w R

Do not use rcpcv.dll (default).

-x str, --config-string=str

Configure TiMidity++ with str.  The format of str is the same as timidity.cfg.

For example:
-x'bank 0\n0 violin.pat'
Sets the instrument number 0 to violin.
Character `\' (Ascii 0x5c) in the str is treated as escape character like in C literals.  For example \n is treated as carriage return.

-Z file, --freq-table=file

Cause the table of frequencies to be read from file.  This is useful to define a tuning different from 12-equal temperament.  If “pure” is specified, TiMidity++ plays in trial pure intonation.

-Zpure[n(m)], --pure-intonation=[n(m)]

Play in trial pure intonation by Key Signature meta-event in the MIDI file.  You can specify the initial keysig by hand, in case the MIDI file does not contains the meta-event.  Optionally, n is the number of key signature.  In case of sharp, n is positive.  In case of flat, n is negative.  Valid values of n are in the interval from -7 to 7.  In case of minor mode, you should put `m' character along with -Zpure option.


Simulate behavior of specific synthesizer module as much as possible. For the moment, the value of n defined is as follows:


TiMidity++ Default


Roland SC-55


Roland SC-88


Roland SC-88Pro


Roland SC-8850


Reserved for GS family










Reserved for XG family


SoundBlaster Live!


SoundBlaster Audigy


Reserved for other synthesizer modules


TiMidity++ Special 1


Reserved for TiMidity++ specification purposes


TiMidity++ Debug

See Also

sf2text(1), timidity.cfg(5)


The latest release is available on the TiMidity++ Page,

URL https://sourceforge.net/projects/timidity/


Eats more CPU time than a small CPU-time-eating animal.

This man page was translated from Japanese to English by me with poor English skill :-)


Version 0.2i and earlier:

Tuukka Toivonen <tt@cgs.fi>
Vincent Pagel <pagel@loria.fr>
Takashi Iwai <tiwai@suse.de>
Davide Moretti <dave@rimini.com>
Chi Ming HUNG <cmhung@insti.physics.sunysb.edu>
Riccardo Facchetti <riccardo@cdc8g5.cdc.polimi.it>


IZUMO Masanao <iz@onicos.co.jp>
HARADA Tomokazu <harada@prince.pe.u-tokyo.ac.jp>
YAMATE Keiichirou <keiich-y@is.aist-nara.ac.jp>
KIRYU Masaki <mkiryu@usa.net>
AOKI Daisuke <dai@y7.net>
MATSUMOTO Shoji <shom@i.h.kyoto-u.ac.jp>
KOYANAGI Masaaki <koyanagi@okilab.oki.co.jp>
IMAI Kunihiko <imai@leo.ec.t.kanazawa-u.ac.jp>
NOGAMI Takaya <t-nogami@happy.email.ne.jp>
WATANABE Takanori <takawata@shidahara1.planet.kobe-u.ac.jp>
TAKEKAWA Hiroshi <sian@big.or.jp>
NAGANO Daisuke <breeze.nagano@nifty.ne.jp>
KINOSHITA kosuke <kino@krhm.jvc-victor.co.jp>
SHIGEMURA Norikatsu <nork@ninth-nine.com>
YAMAHATA Isaku <yamahata@kusm.kyoto-u.ac.jp>
ARAI Yoshishige <ryo2@on.rim.or.jp>
Glenn Trigg <ggt@netspace.net.au>
Tim Allen <thristian@usa.net>
Michael Haardt <michael@moria.de>
Eric A. Welsh <ewelsh@ccb.wustl.edu>
Paolo Bonzini <bonzini@gnu.org>
KIMOTO Masahiko <kimoto@ohnolab.org>
IWAI Takashi <tiwai@suse.de>
Saito <saito2@digitalme.com>
SATO Kentaro <kentaro@ps.catv.ne.jp>
TAMUKI Shoichi <tamuki@linet.gr.jp>
URABE Shohei <root@mput.dip.jp>
SUENAGA Keishi <s_keishi@mutt.freemail.ne.jp>
SUZUKI Koji <k@kekyo.net>

(titles omitted and an order different)

and other many people sends information and bug-fix codes.

The English version of this man page was written by NAGANO Daisuke <breeze.nagano@nifty.ne.jp>.

Now, TAMUKI Shoichi <tamuki@linet.gr.jp> and URABE Shohei <root@mput.dip.jp> are maintaining the man page.
If you have any comments or suggestions or complaints :) about this man page, please tell us it.

Referenced By

midid(1), mscore(1), timidity.cfg(5).

April 25 2004 2.13.0