MPEGplus logo

Frank Klemm's  Musepack  Page

The author of the file format and the original coder and decoder is Andree Buschmann.

His Musepack web page and work you can find at:

Form for bug reports:

        History of this web page
        First Tries of an Encoder
        Decoder Source
        Encoder Source
        Precompiled Decoder Binaries (16 bit output)
        Precompiled Decoder Binaries (16 . . . 32 bit output)
        Precompiled Encoder Binaries
        Monkey's Audio (APE) source and binaries
        Precompiled Test Encoder Binaries
        Audio Links
        WinAMP and XMMS plugins

Other audio information on this site:
        Typical envelope for different K-settings
        Test cases
        Audio file naming proposal
        Kopierschutz (in German)
        Time resolution of different coding technologies
        Clipping of CDs (700 KB images)
        Some remarks about tagging
        Listening Tests
        ECC codes
        Gain controller mess


LATEST:   XMMS plugin (experimental version 0.96)  [New!]

LATEST:   Alpha 1.15f  [New!]

LATEST:   Final 1.1 version  [New!]

Documention of File Format of StreamVersion 8 started

WinAMP tagger + file info preview

Tagger Image

Note: There will be also a simple version which runs at 640x480.

New replaygain calculation tool (0.81a)
Manual (nearly ready for mppdec, started for mppenc):

History of this web page

I've downloaded the original decoder source code (version 1.7.6) in January 2001 and tried to understand the code. The result is an enhanced decoder with improved speed, quality, features, portability and readablity.

Some of these changes were merged back in the original decoder of Andree Buschmann, so also this decoder got a speed improvement of 60% to 100%.

Typical speeds I measured:

Intel 386SX/20 0.0028x Pentium III-866 142x
Intel 386SX/20 + IIT 387SX 0.06x Athlon 700 162x
Motorola 68040/40 0.94x Pentium 4/1.3 184x
Pentium 133 7x Pentium 4/1.7 240x
Pentium 233 MMX 11x Athlon 1333 306x
Pentium II-300 28x Athlon 1400 321x
K6-2-300 43x Pentium 4/3.2 440x

Speed/GHz: i386+ITT: 3, 68040: 24, Pentium: 53, P2: 93, K6-2: 143, P4: 144, P3: 164, Athlon: 232

This shows in the last 10 years CPUs really become faster, although there are a lot of applications tempting to assume the opposite . . .
I have not seen one of AMD new Palomino CPUs so I don't know what is faster: The SSE or the 3DNow! engine.

Advantages of this Decoder:

Features still missing:

Precision (measured with "Brothers in Arms" by Dire Straits):

Samples different from
80 bit calculations
SNR reduction
64 bit C-Code 0 Samples: 0.000 %  0.000 dB
32 bit C-Code 9523 Samples: 0.026 %  0.005 dB
3DNow!/SSE-Code 12358 Samples: 0.033 %  0.006 dB
1.7.6a version 49.826 %  6.005 dB
1.7.8a version 9528 Samples: 0.026 %  0.006 dB

Samples always differ by only +/- 1.

Note: Recommentations and bug reports are welcomed (use a realname and a valid e-Mail address!).

First Tries of an Encoder

Typical speeds I measured:

Intel 386SX/20 0.00024x Pentium III-850 6.0x
Intel 386SX/20 + IIT 387SX ? Athlon 700 7.8x
Pentium 133 ? Pentium 4/1.3 8.7x
Pentium 233 MMX ? Pentium 4/1.7 11.7x
Pentium II-300 2.45x Athlon 1333 14.6x
K6-2-300 ? Athlon 1400 15.4x
Celeron 500 3.6x
K6-2-400 1.0x K6-III-420 1.5x

Speed/GHz: i386: 0.012, P4: 6.8, Celeron: 7.2, P2: 8.2, P3: ca.8, Athlon: 10.9

Advantages of this Encoder:

Still missing:

Note: Recommentations and bug reports are welcomed (use a realname and a valid e-Mail address!).

Decoder Source:

Note (only for Intel x86):  For maximum speed you need NASM, the Netwide Assembler.
Otherwise you must translate without NASM support (#undef USE_ASM).

Current release version: 1.1, 2002-03-25  (ISO 8601).

Edit mpp.h and select the features which are possible or you need.
Edit the Makefile to add compiler options and libraries you need.


XMMS source

Encoder Source:

PGP encrypted for selected persons

Current encoder source: 1.1, 2002-03-25  (ISO 8601).

Current source: 1.93, 2002-10-25  (ISO 8601).

Precompiled Decoder Binaries (16 bit output):  (zipped with zip/pkzip/winzip)

Download, unzip and copy the file to a location in your path (usually /usr/local/bin or /usr/bin).
Set the owner to root and the rights to 4755 if you wish to use a higher priority of the program for audio playback (only UNIX).

Currently I'm transforming als binaries to the notation:



Other Operating Systems:

I'm still searching for compiled and tested executables for:
Test cases (taken from

Precompiled Decoder Binaries (16 . . . 32 bit output):  (zipped with zip/pkzip/winzip)

Other Operating Systems:


How to roll your own:

Other remarks:

Precompiled Encoder Binaries:  (zipped with zip/pkzip/winzip)

Download, unzip and copy the file to a location in your path (usually /usr/local/bin or /usr/bin).
Set the owner to root and the rights to 4755 if you wish to use a idle or higher priority of the program (only Unix).


Other Operating Systems:

Monkeys Audio (APE):   [Monkey]

Precompiled Test Encoder Binaries:  (zipped with zip/pkzip/winzip)

The purpose for Test Encoder Binaries are listening tests and NOT FOR ARCHIVAL.

The file format can change every day, so for testing use the newest decoder source and compile it with StreamVersion 8 support on.

The purpose of StreamVersion 8 is not to reduce the typical bitrate, but to increase quality, robustness and handling. Typical bitrates will always around 170 kbps for transparent encoding.


Other Operating Systems:


Purposes: You need:

Linux: Windows: Other OS: How does it works: Bugs: No bugs are: Examples: (note that Microsoft's Compiler currently only supports one indirection, Unix multiple until a certain degree)

Audio Links:

General MPEG-1/2 Layer 1 (.mp1) MPEG-1/2 Layer 2 (.mp2) .mpc Musepack / MPEGplus (.mp+ / .mpc) MPEG-1/2 Layer 3 (.mp3) MPEG-2/4 Advanced Audio Coding (.aac) MPEG-4 TwinVQ (.vqf) MPEG-2 Video / MPEG-2 System Stream (.mpeg) MPEG-4 Video / MPEG-4 System Stream (.mpeg) Ogg Vorbis (.ogg) Dolby Digital (.ac3) Dolby Surround EX Digital Theater Systems (.dts) DTS ES DTS ES discrete Sony Dynamic Digital Sound (.sdds) Adaptive Transform Acoustic Coding 1 (ATRAC) (used on Minidisk) Adaptive Transform Acoustic Coding 2/3 (ATRAC3) (used on Minidisk LP) Low bitrate and mobile speech coding Ogg / Speex

Norm bitrate (kbps) coding used by
ITU-T G.711 64 non linear PCM
(A-law / µ-law)
ITU-T G.726 16, 24, 32, 40 ADPCM DCME
ITU-T G.723.1 5.3
Voice over IP
ITU-T G.728 16 LD-CELP Long distance telephone
ITU-T G.729 8 CS-ACELP Voice over IP / Digital satellite / HQ digital radio
full rate
13 RPE-LTP European Mobile telephony
half rate
5.6 VSELP European Mobile telephony
enhanced full rate
13 ACELP European Mobile telephony (1.8 GHz band)

MPEG-4 Sound Field Synthesis [Studio]

Spectral Band Replication (SBR)

SBR is a technique to virtually enhance the frequency response by "guessing" the upper audio signal by the lower audio signal and some additional data bits. See There are some hints that it is like that this becomes a future MPEG standard, see here.

Examples: Other SBR technique: Still remaining: ePAC, RM, MOV, WMA, ASF, . . .

Lossless Audio Compression Digital Audio Recording Equipment (black living room 17" / 43 cm equipment)
    DVD-5       (12 cm, SS/SL)   4.37  gig ( 4.70  BB)
    DVD-9       (12 cm, SS/DL)   7.95  gig ( 8.54  BB)
    DVD-10      (12 cm, DS/SL)   8.74  gig ( 9.40  BB)
    DVD-14      (12 cm, DS/ML)  12.32  gig (13.24  BB)
    DVD-18      (12 cm, DS/DL)  15.90  gig (17.08  BB)
    DVD-1       ( 8 cm, SS/SL)   1.36  gig ( 1.46  BB)
    DVD-2       ( 8 cm, SS/DL)   2.47  gig ( 2.66  BB)
    DVD-3       ( 8 cm, DS/SL)   2.72  gig ( 2.92  BB)
    DVD-4       ( 8 cm, DS/DL)   4.95  gig ( 5.32  BB)
    DVD-R 1.0   (12 cm, SS/SL)   3.68  gig ( 3.95  BB)
    DVD-R 2.0   (12 cm, SS/SL)   4.37  gig ( 4.70  BB)
    DVD-R 2.0   (12 cm, DS/SL)   8.75  gig ( 9.40  BB)  rare!
    DVD-RW 2.0  (12 cm, SS/SL)   4.37  gig ( 4.70  BB)
    DVD-RW 2.0  (12 cm, DS/SL)   8.75  gig ( 9.40  BB)  rare!
    DVD-RAM 1.0 (12 cm, SS/SL)   2.40  gig ( 2.58  BB)
    DVD-RAM 1.0 (12 cm, DS/SL)   4.80  gig ( 5.16  BB)
    DVD-RAM 2.0 (12 cm, SS/SL)   4.37  gig ( 4.70  BB)
    DVD-RAM 2.0 (12 cm, DS/SL)   8.75  gig ( 9.40  BB)
    DVD-RAM 2.0 ( 8 cm, SS/SL)   1.36  gig ( 1.46  BB)
    DVD-RAM 2.0 ( 8 cm, DS/SL)   2.47  gig ( 2.65  BB)
    CD-ROM      (12 cm, SS/SL)   0.635 gig ( 0.650 BB)
    CD-ROM      ( 8 cm, SS/SL)   0.180 gig ( 0.194 BB)
    DDCD-ROM    (12 cm, SS/SL)   1.270 gig ( 1.364 BB)
    DDCD-ROM    ( 8 cm, SS/SL)   0.360 gig ( 0.387 BB)

Note, there are also other digital audio formats, but they are not high quality (example: VTR Hi8 digital audio)

WinAMP and XMMS plugins

WinAMP Input-Plugins: WinAmp XMMS Input-Plugins: XMMS

CD/DVD sellers + Information (German):

Room Acostic:

Disclaimer:  (in German)

Mit Urteil vom 12. Mai 1998 hat das Landgericht Hamburg entschieden, daß man durch die Ausbringung eines Links die Inhalte der gelinkten Seite ggf. mit zu verantworten hat. Dies kann - so das Landgericht - nur dadurch verhindert werden, daß man sich ausdrücklich von diesen Inhalten distanziert.

Für alle auf diesen Seiten aufgeführten Links gilt:

Ich möchte ausdrücklich betonen, daß ich keinerlei Einfluß auf die Gestaltung und die Inhalte der gelinkten Seiten habe. Deshalb distanziere ich mich hiermit ausdrücklich von allen Inhalten aller gelinkten Seiten auf meiner Homepage und mache mir deren Inhalte nicht zu eigen. Diese Erklärung gilt für alle auf meiner Homepage ausgebrachten Links und für alle Inhalte der Seiten, zu denen die Banner und Links führen.

Last modified:  2002-05-27                                Visitors:  

[eMail]      [Addr] . . . . .