Swift ReplayKit AVAssetWriter Video Audio out of Sync when Converted to HLS - ios

Question

In iOS/Swift I am working with ReplayKit to use AVAssetWriter to create a mov or MP4 video of the user's screen and microphone audio.
When I create a video, it plays fine locally and the audio and video are in sync. However when I convert this video to HLS (HTTP Live Stream) format using AWS Mediaconvert, the audio is out of sync with the video. Does anyone know what could be causing this? I read about timecoding, maybe I need to add a timecode to my video? Is there an easier way to fix this or has anyone experience similar issues?
private func startRecordingVideo(){
//Initialize MP4 Output File for Screen Recorded Video
let fileManager = FileManager.default
let urls = fileManager.urls(for: .documentDirectory, in: .userDomainMask)
guard let documentDirectory: NSURL = urls.first as NSURL? else {
fatalError("documentDir Error")
}
videoOutputURL = documentDirectory.appendingPathComponent("OutputVideo.mov")
if FileManager.default.fileExists(atPath: videoOutputURL!.path) {
do {
try FileManager.default.removeItem(atPath: videoOutputURL!.path)
} catch {
fatalError("Unable to delete file: \(error) : \(#function).")
}
}
//Initialize Asset Writer to Write Video to User's Storage
assetWriter = try! AVAssetWriter(outputURL: videoOutputURL!, fileType:
AVFileType.mov)
let videoOutputSettings: Dictionary<String, Any> = [
AVVideoCodecKey : AVVideoCodecType.h264,
AVVideoWidthKey : UIScreen.main.bounds.size.width,
AVVideoHeightKey : UIScreen.main.bounds.size.height,
];
let audioSettings = [
AVFormatIDKey : kAudioFormatMPEG4AAC,
AVNumberOfChannelsKey : 1,
AVSampleRateKey : 44100.0,
AVEncoderBitRateKey: 96000,
] as [String : Any]
videoInput = AVAssetWriterInput(mediaType: AVMediaType.video,outputSettings: videoOutputSettings)
audioInput = AVAssetWriterInput(mediaType: AVMediaType.audio,outputSettings:audioSettings )
videoInput?.expectsMediaDataInRealTime = true
audioInput?.expectsMediaDataInRealTime = true
assetWriter?.add(videoInput!)
assetWriter?.add(audioInput!)
let sharedRecorder = RPScreenRecorder.shared()
sharedRecorder.isMicrophoneEnabled = true
sharedRecorder.startCapture(handler: {
(sample, bufferType, error) in
//Audio/Video Buffer Data returned from the Screen Recorder
if CMSampleBufferDataIsReady(sample) {
DispatchQueue.main.async { [weak self] in
//Start the Asset Writer if it has not yet started
if self?.assetWriter?.status == AVAssetWriter.Status.unknown {
if !(self?.assetWriter?.startWriting())! {
return
}
self?.assetWriter?.startSession(atSourceTime: CMSampleBufferGetPresentationTimeStamp(sample))
self?.startSession = true
}
}
//Handle errors
if self.assetWriter?.status == AVAssetWriter.Status.failed {
print("Error occured, status = \(String(describing: self.assetWriter?.status.rawValue)), \(String(describing: self.assetWriter?.error!.localizedDescription)) \(String(describing: self.assetWriter?.error))")
return
}
//Add video buffer to AVAssetWriter Video Input
if (bufferType == .video)
{
if(self.videoInput!.isReadyForMoreMediaData) && self.startSession {
self.videoInput?.append(sample)
}
}
//Add audio microphone buffer to AVAssetWriter Audio Input
if (bufferType == .audioMic)
{
print("MIC BUFFER RECEIVED")
if self.audioInput!.isReadyForMoreMediaData
{
print("Audio Buffer Came")
self.audioInput?.append(sample)
}
}
}
}, completionHandler: {
error in
print("COMP HANDLER ERROR", error?.localizedDescription)
})
}
private func stopRecordingVideo(){
self.startSession = false
RPScreenRecorder.shared().stopCapture{ (error) in
self.videoInput?.markAsFinished()
self.audioInput?.markAsFinished()
if error == nil{
self.assetWriter?.finishWriting{
self.startSession = false
print("FINISHED WRITING!")
DispatchQueue.main.async {
self.setUpVideoPreview()
}
}
}else{
//DELETE DIRECTORY
}
}
}

Answer 1

I’m sure you’ve either figured this out or moved on, but for all Googlers you basically have to set the mediaTimeScale on the video input. You can see an example here
Here’s the relevant part of that code (This code is using a AVSampleBufferDisplayLayer, but the same concept applies:
double pts = CMTimeGetSeconds(CMSampleBufferGetPresentationTimeStamp(sampleBuffer));
if(!timebaseSet && pts != 0)
{
timebaseSet = true;
CMTimebaseRef controlTimebase;
CMTimebaseCreateWithMasterClock( CFAllocatorGetDefault(), CMClockGetHostTimeClock(), &controlTimebase );
displayLayer.controlTimebase = controlTimebase;
CMTimebaseSetTime(displayLayer.controlTimebase, CMTimeMake(pts, 1));
CMTimebaseSetRate(displayLayer.controlTimebase, 1.0);
}
if([displayLayer isReadyForMoreMediaData])
{
[displayLayer enqueueSampleBuffer:sampleBuffer];
}