Porting LFLiveKit to macOS
LFLiveKit is an open source RTMP streaming SDK for iOS, and the project ports LFLiveKit to macOS. It helps you to broadcast RTMP streams in your macOS apps.
Currently it is only tested on
- macOS 10.13 and Xcode 9.1 to 9.2.
- macOS 10.12 and Xcode 8.3.
However, it should support Xcode 7 and macOS 10.8 and above.
Unlike iOS, it is very possible that a mac user has a lot of external audio and video capture devices connected, such as USB microphones, webcams and so on. Thus, we change the interface for creating LFLiveSession objects, specifying a video device and an audio device is required.
You may have code like this:
LFLiveAudioConfiguration *audioConfig = [LFLiveAudioConfiguration defaultConfiguration];
LFLiveVideoConfiguration *videoConfig = [LFLiveVideoConfiguration defaultConfigurationForQuality:LFLiveVideoQuality_Medium3];
AVCaptureDevice *audioDevice = [LFLiveSession availableAudioDevices][0];
AVCaptureDevice *videoDevice = [LFLiveSession availableCameraDevices][0];
_session = [[LFLiveSession alloc] initWithAudioConfiguration:audioConfig audioDevice:audioDevice videoConfiguration:videoConfig videoDevice:videoDevice captureType:LFLiveCaptureMaskAll];
_session.delegate = self;The SDK supports CocoaPods. Please add pod LFLiveKitMac
to your Podfile, and then call pod install.
Since APIs differ between iOS and macOS, we need to modify LFLiveKit to make it able to build and run on macOS. What we did are including
- Change all of the OpenES API calls to corresponding OpenGL ones.
- Change the audio component for audio recording. LFLiveKit uses the remoteIO node on iOS, but we need to use kAudioUnitSubType_HALOutput on mac OS.
- Change the way that GPUImageFramebuffer generates pixel buffer. GPUImageFramebuffer uses an iOS specific texture cache to read processed textures and write them into CVPixelBufferRef objects. However, the cache does not exist on macOS. We have to use another way to create CVPixelBufferRef objects.
- LFHardwareAudioEncoder does not work on macOS, so we always use LFH264VideoEncoder.
- Some video filters are removes temporarily in order to make the project simpler while doing porting.
- Settings for camera torch and rotation are removed as well, since macOS does not support them.
If you try to obtain camera and audio devices by calling
[LFLiveSession availableAudioDevices] and
[LFLiveSession availableCameraDevices], and they return nothing at
all, you might want to check the "Capabilities" tab in Xcode.
Please select your application target in your Xcode project, and select the second tab "Capabilities". Then, click on the check boxes of "Hardware: Camera" and "Hardware: Microphone". You are not able to access users' hardware until completing the settings.
Apple provides an API, the Audio Power Hint, which helps saving
battery life. Once you want to help users to save more power, you can
set kAudioHardwarePropertyPowerHint to
kAudioHardwarePowerHintFavorSavingPower by calling
AudioObjectSetPropertyData, or set AudioHardwarePowerHint key to
Favor Saving Power in your Info.plist file.
Set Audio Power Hint to saving power is great for audio playback, but it is horrible for recording. When you are in power saving mode, the I/O buffer size is changed and it may cause what you are recording to become noises.
For further information about Audio Power Hint, please take a look at Technical Note TN2321.
Enjoy! 😀
LFLiveKit is released under the MIT license. See LICENSE for details.