SoundTouch audio processing library v1.5.0

SoundTouch library Copyright (c) Olli Parviainen 2002-2009

1. Introduction

SoundTouch is an open-source audio processing library that allows changing the sound tempo, pitch and playback rate parameters independently from each other, i.e.:

1.1 Contact information

Author email: oparviai 'at' iki.fi

SoundTouch WWW page: http://www.surina.net/soundtouch

2. Compiling SoundTouch

Before compiling, notice that you can choose the sample data format if it's desirable to use floating point sample data instead of 16bit integers. See section "sample data format" for more information.

2.1. Building in Microsoft Windows

Project files for Microsoft Visual C++ 6.0 and Visual C++ .NET are supplied with the source code package. 

Please notice that SoundTouch library uses processor-specific optimizations for Pentium III and AMD processors. Visual Studio .NET and later versions supports the required instructions by default, but Visual Studio 6.0 requires a processor pack upgrade to be installed in order to support these optimizations. The processor pack upgrade can be downloaded from Microsoft site at this URL:

http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx

If the above URL is unavailable or removed, go to http://msdn.microsoft.com and perform a search with keywords "processor pack".

To build the binaries with Visual C++ compiler, either run "make-win.bat" script, or open the appropriate project files in source code directories with Visual Studio. The final executable will appear under the "SoundTouch\bin" directory. If using the Visual Studio IDE instead of the make-win.bat script, directories bin and lib may need to be created manually to the SoundTouch package root for the final executables. The make-win.bat script creates these directories automatically.

2.2. Building in Gnu platforms

The SoundTouch library can be compiled in practically any platform supporting GNU compiler (GCC) tools. SoundTouch have been tested with gcc version 3.3.4., but it shouldn't be very specific about the gcc version. Assembler-level performance optimizations for GNU platform are currently available in x86 platforms only, they are automatically disabled and replaced with standard C routines in other processor platforms.

To build and install the binaries, run the following commands in the SoundTouch/ directory:

./configure  -

Configures the SoundTouch package for the local environment.

 
make         -

Builds the SoundTouch library & SoundStretch utility.

 
make install -

Installs the SoundTouch & BPM libraries to /usr/local/lib and SoundStretch utility to /usr/local/bin. Please notice that 'root' privileges may be required to install the binaries to the destination locations.

2.2.1 Required GNU tools 

Bash shell, GNU C++ compiler, libtool, autoconf and automake tools are required for compiling the SoundTouch library. These are usually included with the GNU/Linux distribution, but if not, install these packages first. For example, in Ubuntu Linux these can be acquired and installed with the following command:

sudo apt-get install automake autoconf libtool build-essential

2.2.2 Problems with GCC compiler compatibility

At the release time the SoundTouch package has been tested to compile in GNU/Linux platform. However, in past it's happened that new gcc versions aren't necessarily compatible with the assembler settings used in the optimized routines. If you have problems getting the SoundTouch library compiled, try the workaround of disabling the optimizations by editing the file "include/STTypes.h" and removing the following definition there:

#define ALLOW_OPTIMIZATIONS 1

2.2.3 Problems with configure script or build process 

Incompatibilities between various GNU toolchain versions may cause errors when running the "configure" script or building the source codes, if your GNU tool versions are not compatible with the versions used for preparing the SoundTouch kit. 

To resolve the issue, regenerate the configure scripts with your local tool set by running the "./bootstrap" script included in the SoundTouch source code kit. After that, run the configure script and make as usually.

2.2.4 Compiler issues with non-x86 processors

SoundTouch library works also on non-x86 processors.

However, in case that you get compiler errors when trying to compile for non-Intel processor, edit the file "source\SoundTouch\Makefile.am" and remove the "-msse2" flag on the AM_CXXFLAGS line:

AM_CXXFLAGS=-O3 -fcheck-new -I../../include    # Note: -msse2 flag removed!

After that, run "./bootstrap" script, and then run configure and make again.

3. About implementation & Usage tips

3.1. Supported sample data formats

The sample data format can be chosen between 16bit signed integer and 32bit floating point values, the default is 32bit floating point.

In Windows environment, the sample data format is chosen in file "STTypes.h" by choosing one of the following defines:

In GNU environment, the floating sample format is used by default, but integer sample format can be chosen by giving the following switch to the configure script: 

./configure --enable-integer-samples

The sample data can have either single (mono) or double (stereo) audio channel. Stereo data is interleaved so that every other data value is for left channel and every second for right channel. Notice that while it'd be possible in theory to process stereo sound as two separate mono channels, this isn't recommended because processing the channels separately would result in losing the phase coherency between the channels, which consequently would ruin the stereo effect.

Sample rates between 8000-48000H are supported.

3.2. Processing latency

The processing and latency constraints of the SoundTouch library are:

3.3. About algorithms

SoundTouch provides three seemingly independent effects: tempo, pitch and playback rate control. These three controls are implemented as combination of two primary effects, sample rate transposing and time-stretching.

Sample rate transposing affects both the audio stream duration and pitch. It's implemented simply by converting the original audio sample stream to the  desired duration by interpolating from the original audio samples. In SoundTouch, linear interpolation with anti-alias filtering is used. Theoretically a higher-order interpolation provide better result than 1st order linear interpolation, but in audio application linear interpolation together with anti-alias filtering performs subjectively about as well as higher-order filtering would.

Time-stretching means changing the audio stream duration without affecting it's pitch. SoundTouch uses WSOLA-like time-stretching routines that operate in the time domain. Compared to sample rate transposing, time-stretching is a much heavier operation and also requires a longer processing "window" of sound samples used by the processing algorithm, thus increasing the algorithm input/output latency. Typical i/o latency for the SoundTouch time-stretch algorithm is around 100 ms.

Sample rate transposing and time-stretching are then used together to produce the tempo, pitch and rate controls:

3.4 Tuning the algorithm parameters

The time-stretch algorithm has few parameters that can be tuned to optimize sound quality for certain application. The current default parameters have been chosen by iterative if-then analysis (read: "trial and error") to obtain best subjective sound quality in pop/rock music processing, but in applications processing different kind of sound the default parameter set may result into a sub-optimal result.

The time-stretch algorithm default parameter values are set by the following #defines in file "TDStretch.h":

#define DEFAULT_SEQUENCE_MS     AUTOMATIC
#define DEFAULT_SEEKWINDOW_MS   AUTOMATIC
#define DEFAULT_OVERLAP_MS      8

These parameters affect to the time-stretch algorithm as follows:

Notice that these parameters can also be set during execution time with functions "TDStretch::setParameters()" and "SoundTouch::setSetting()".

The table below summaries how the parameters can be adjusted for different applications:

Parameter name Default value magnitude Larger value affects... Smaller value affects... Effect to CPU burden 
SEQUENCE_MS
 Default value is relatively large, chosen for slowing down music tempo Larger value is usually better for slowing down tempo. Growing the value decelerates the "echoing" artifact when slowing down the tempo. Smaller value might be better for speeding up tempo. Reducing the value accelerates the "echoing" artifact when slowing down the tempo Increasing the parameter value reduces computation burden 
SEEKWINDOW_MS
 Default value is relatively large, chosen for slowing down music tempo Larger value eases finding a good mixing position, but may cause a "drifting" artifact Smaller reduce possibility to find a good mixing position, but reduce the "drifting" artifact. Increasing the parameter value increases computation burden 
OVERLAP_MS
 Default value is relatively large, chosen to suit with above parameters.   If you reduce the "sequence ms" setting, you might wish to try a smaller value. Increasing the parameter value increases computation burden

3.5 Performance Optimizations

General optimizations:

The time-stretch routine has a 'quick' mode that substantially speeds up the algorithm but may degrade the sound quality by a small amount. This mode is activated by calling SoundTouch::setSetting() function with parameter  id of SETTING_USE_QUICKSEEK and value "1", i.e.

setSetting(SETTING_USE_QUICKSEEK, 1);

CPU-specific optimizations:

4. SoundStretch audio processing utility

SoundStretch audio processing utility
Copyright (c) Olli Parviainen 2002-2009

SoundStretch is a simple command-line application that can change tempo, pitch and playback rates of WAV sound files. This program is intended primarily to demonstrate how the "SoundTouch" library can be used to process sound in your own program, but it can as well be used for processing sound files.

4.1. SoundStretch Usage Instructions

SoundStretch Usage syntax:

soundstretch infilename outfilename [switches]

Where: 

"infilename"
 Name of the input sound data file (in .WAV audio file format). Give "stdin" as filename to use standard input pipe. 
"outfilename"
 Name of the output sound file where the resulting sound is saved (in .WAV audio file format). This parameter may be omitted if you  don't want to save the output (e.g. when only calculating BPM rate with '-bpm' switch). Give "stdout" as filename to use standard output pipe. 
 [switches]
 Are one or more control switches.

Available control switches are:

-tempo=n
Change the sound tempo by n percents (n = -95.0 .. +5000.0 %) 
-pitch=n
 Change the sound pitch by n semitones (n = -60.0 .. + 60.0 semitones) 
-rate=n
 Change the sound playback rate by n percents (n = -95.0 .. +5000.0 %) 
-bpm=n
 Detect the Beats-Per-Minute (BPM) rate of the sound and adjust the tempo to meet 'n' BPMs. When this switch is applied, the "-tempo" switch is ignored. If "=n" is omitted, i.e. switch "-bpm" is used alone, then the BPM rate is estimated and displayed, but tempo not adjusted according to the BPM value. 
-quick
 Use quicker tempo change algorithm. Gains speed but loses sound quality. 
-naa
 Don't use anti-alias filtering in sample rate transposing. Gains speed but loses sound quality. 
-license
 Displays the program license text (LGPL)

Notes:

4.2. SoundStretch usage examples

Example 1

The following command increases tempo of the sound file "originalfile.wav" by 12.5% and stores result to file "destinationfile.wav":

soundstretch originalfile.wav destinationfile.wav -tempo=12.5

Example 2

The following command decreases the sound pitch (key) of the sound file "orig.wav" by two semitones and stores the result to file "dest.wav":

soundstretch orig.wav dest.wav -pitch=-2

Example 3

The following command processes the file "orig.wav" by decreasing the sound tempo by 25.3% and increasing the sound pitch (key) by 1.5 semitones. Resulting .wav audio data is directed to standard output pipe:

soundstretch orig.wav stdout -tempo=-25.3 -pitch=1.5

Example 4

The following command detects the BPM rate of the file "orig.wav" and adjusts the tempo to match 100 beats per minute. Result is stored to file "dest.wav":

soundstretch orig.wav dest.wav -bpm=100

Example 5

The following command reads .wav sound data from standard input pipe and estimates the BPM rate:

soundstretch stdin -bpm

5. Change History

5.1. SoundTouch library Change History

1.5.0:

1.4.1:

1.4.0:

v1.3.1:

v1.3.0:

v1.2.1:

v1.2.0:

v1.1.1:

v1.01:

v1.0:

 

5.2. SoundStretch application Change History

1.4.0:

v1.3.0:

v1.2.1:

v1.2.0:

v1.1.1:

v1.1:

v1.01:

6. Acknowledgements

Kudos for these people who have contributed to development or submitted bugfixes since SoundTouch v1.3.1:

Moral greetings to all other contributors and users also!

7. LICENSE

SoundTouch audio processing library
Copyright (c) Olli Parviainen

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License version 2.1 as published by the Free Software Foundation.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with this library; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA