0bmxa / Pancake

macOS audio server plug-in framework

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool



Pancake is a framework for writing macOS AudioServer (coreaudiod) plug-ins for realtime audio processing in the system audio output path.

It is written in Swift and provides highly simplified* Swift and C APIs for setting up your own systemwide audio processing solution.
*in comparison to Apple's audio server plugin API

Getting Started

Clone the project and check out the sample driver. The usage of the framework is as simple as the following 4 steps:

// Configure a virtual device
CFStringRef manufacturer, name, UID = ...;
PancakeDeviceConfiguration *device = CreatePancakeDeviceConfig(manufacturer, name, UID);

// Add a supported audio format
AudioStreamBasicDescription supportedFormat = ...;
PancakeDeviceConfigAddFormat(device, supportedFormat);

// Create a pancake config & add the device
PancakeConfiguration *config = CreatePancakeConfig(NULL);
PancakeConfigAddDevice(config, device);

// Setup Pancake with the config

How it works

Pancake consists of three components:

  1. An audio server plugin
  2. An audio routing helper (background) app
  3. A settings app

The audio server plugin creates a virtual audio device, which provides an audio output to the system, just like any regular audio output (like Headphones, AirPlay, BT, etc.). Any audio that is sent to this output is processed by a (custom providable) processing callback and routed back to the system on a virtual audio input.

To make the processing transparent to the user, those inputs & outputs can be hidden from the UI. In this case the audio routing helper overwrites the system output in the background to the (hidden) virtual one, and also routes the virtual input to the actual, user-selected audio output.

As both, the plugin and the helper, are not visible to the user, the settings app provides an interface to the user, where they can change parameters, which are forwarded to the plugin and/or helper.



The project heavily uses Swift 4, so an up-to-date Swift toolchain is required. It also includes some C (and Objective C) parts, so also clang is required. The easiest way to get all of this at once is by installing the latest Xcode from the Mac App Store.

mas install 497799835

The project also uses Google's cpplint and Realm's SwiftLint for linting, so make sure you have both of them installed and in your PATH

# Install SwiftLint via homebrew
brew install swiftlint

# Manually download cpplint.py to a dir of your choice & make it executable
curl 'https://raw.githubusercontent.com/google/styleguide/gh-pages/cpplint/cpplint.py' \
    -o '/usr/local/bin/cpplint.py'
chmod +x '/usr/local/bin/cpplint.py'


The projects creates an audio server plugin (.driver bundle), which has to be placed in /Library/Audio/Plug-Ins/HAL/ for the audio server being able to find and load it. When building the project via Xcode, the plugin is automatically copied to this directory.

For Xcode being able to write to /Library/Audio/Plug-Ins/HAL/, the directory needs to be writable by the current user. The repo includes an AppleScript file CheckAndFixInstallPermissions.scpt which attempts to check and fix this for you.

Alternatively you can just manually run:

sudo chmod o+w /Library/Audio/Plug-Ins/HAL/

Please note that making a system directory writeable might cause a security risk!

If you don't want to change permissions on the directory, you can also copy the build product manually to /Library/Audio/Plug-Ins/HAL/--this, however, makes debugging a lot harder.


For the plugin to be loaded by the audio server, the audio server (coreaudiod) has to be restarted. This can be triggered by just killing coreaudiod, or by gracefully restarting the launchd job:

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.audio.coreaudiod.plist
sudo launchctl load /System/Library/LaunchDaemons/com.apple.audio.coreaudiod.plist

As on restart of the audio server, all audio client connections will be terminated, you might want to restart some system components as well, like

killall SystemUIServer # The menu bar
killall ControlStrip   # The MBP TouchBar, if you have one


For debugging, you need to attach your debugger to coreaudiod, which loads and executes the plugin. You can simply run your own instance from /usr/sbin/coreaudiod and get the basic stuff running.

However, the daemon's activity is quite limited when being run as a seperate instance. So for proper debugging you need to attach to the OS's process, which sadly (actually luckily, but not here) is protected from being attached to by macOS' System Integrity Protection.

⚠️ Disabling System Integrity Protection can cause serious security issues. Do this at your own risk and only if you understand the consequences! ⚠️

Luckily System Integrity Protection is modular and the corresponding csrutil has some nice undocumented options, so you don't have to disable whole SIP. See here for more info about those options: Apple Developer Forums: Thread 17452.


We use SemVer for versioning. For the versions available, see the tags on this repository.


  • Max Heimwrote his master thesis about the audio HAL and created the first version of this framework.

See also the list of contributors who participated in this project.


This project is licensed under the Apache License 2.0 - see the LICENSE.md file for details


  • Hat tip to anyone who's code was used
  • Inspiration
  • etc


macOS audio server plug-in framework

License:Apache License 2.0


Language:Swift 88.6%Language:Objective-C++ 3.8%Language:C++ 3.1%Language:Objective-C 2.7%Language:Ruby 1.0%Language:C 0.7%Language:Shell 0.1%