This is the codebase for Ableton Link, a technology that synchronizes musical beat, tempo, and phase across multiple applications running on one or more devices. Applications on devices connected to a local network discover each other automatically and form a musical session in which each participant can perform independently: anyone can start or stop while still staying in time. Anyone can change the tempo, the others will follow. Anyone can join or leave without disrupting the session.
Ableton Link is dual licensed under GPLv2+ and a proprietary license. If you would like to incorporate Link into a proprietary software application, please contact [email protected].
Link relies on asio-standalone and catch as submodules. After checking out the
main repositories, those submodules have to be loaded using
git submodule update --init --recursive
Link uses CMake to generate build files for the Catch-based unit-tests and the example applications.
$ mkdir build
$ cd build
$ cmake ..
$ cmake --build .
The output binaries for the example applications and the unit-tests will be placed in a
bin subdirectory of the CMake binary directory.
To make sure users have the best possible experience using Link it is important all apps supporting Link behave consistently. This includes for example playing in sync with other apps as well as not hijacking a jams tempo when joining. To make sure your app behaves as intended make sure it complies to the Test Plan.
Link is a header-only library, so it should be straightforward to integrate into your application.
If you are using CMake, then you can simply add the following to your CMakeLists.txt file:
include($PATH_TO_LINK/AbletonLinkConfig.cmake)
target_link_libraries($YOUR_TARGET Ableton::Link)
You can optionally have your build target depend on ${link_HEADERS}, which will make
the Link headers visible in your IDE. This variable exported to the PARENT_SCOPE by
Link's CMakeLists.txt.
To include the Link library in your non CMake project, you must do the following:
- Add the
link/includeandmodules/asio-standalone/asio/includedirectories to your list of include paths - Define
LINK_PLATFORM_MACOSX=1,LINK_PLATFORM_LINUX=1, orLINK_PLATFORM_WINDOWS=1, depending on which platform you are building on.
If you get any compiler errors/warnings, have a look at compile-flags.cmake, which might provide some insight as to the compiler flags needed to build Link.
| Platform | Minimum Required | Optional (only required for examples) |
|---|---|---|
| Windows | MSVC 2013 | Steinberg ASIO SDK 2.3 |
| Mac | Xcode 7.0 | |
| Linux | Clang 3.6 or GCC 5.2 | libportaudio19-dev |
Other compilers with good C++11 support should work, but are not verified.
iOS developers should not use this repo. See http://ableton.github.io/linkkit for information on the LinkKit SDK for iOS.
An overview of Link concepts can be found at http://ableton.github.io/link. Those that
are new to Link should start there. The Link.hpp header
contains the full Link public interface. See the LinkHut projects in this repo for an
example usage of the Link type.
Link works by calculating a relationship between the system clocks of devices in a session.
Since the mechanism for obtaining a system time value and the unit of these values differ
across platforms, Link defines a Clock abstraction with platform-specific
implementations. Please see:
Link::clock()method in Link.hpp- OSX and iOS clock implementation in platforms/darwin/Clock.hpp
- Windows clock implementation in platforms/windows/Clock.hpp
- C++ standard library
std::chrono::high_resolution_clock-based implementation in platforms/stl/Clock.hpp
Using the system time correctly in the context of an audio callback gets a little complicated. Audio devices generally have a sample clock that is independent of the system Clock. Link maintains a mapping between system time and beat time and therefore can't use the sample time provided by the audio system directly.
On OSX and iOS, the CoreAudio render callback is passed an AudioTimeStamp structure with
a mHostTime member that represents the system time at which the audio buffer will be
passed to the audio hardware. This is precisely the information needed to derive the beat
time values corresponding to samples in the buffer using Link. Unfortunately, not all
platforms provide this data to the audio callback.
When a system timestamp is not provided with the audio buffer, the best a client can do in the audio callback is to get the current system time and filter it based on the provided sample time. Filtering is necessary because the audio callback will not be invoked at a perfectly regular interval and therefore the queried system time will exhibit jitter relative to the sample clock. The Link library provides a HostTimeFilter utility class that performs a linear regression between system time and sample time in order to improve the accuracy of system time values used in an audio callback. See the audio callback implementations for the various platforms used in the examples to see how this is used in practice. Note that for Windows-based systems, we recommend using the ASIO audio driver.
As discussed in the previous section, the system time that a client is provided in an audio callback either represents the time at which the buffer will be submitted to the audio hardware (for OSX/iOS) or the time at which the callback was invoked (when the code in the callback queries the system time). Note that neither of these is what we actually want to synchronize between devices in order to play in time.
In order for multiple devices to play in time, we need to synchronize the moment at which their signals hit the speaker or output cable. If this compensation is not performed, the output signals from devices with different output latencies will exhibit a persistent offset from each other. For this reason, the audio system's output latency should be added to system time values before passing them to Link methods. Examples of this latency compensation can be found in the platform implementations of the example apps.