how to merge video clips using avfoundation? - ios

Question

I have successfully merge the video clips to a single video but I am having a problem in the final merged video, the final video shows a white frame after the end of every video clip. I have tried a lot to remove this but couldn't find success. Please review my code below.
func merge(arrayVideos:[AVAsset], completion:#escaping (_ exporter: AVAssetExportSession) -> ()) -> Void {
let mainComposition = AVMutableComposition()
let compositionVideoTrack = mainComposition.addMutableTrack(withMediaType: .video, preferredTrackID: kCMPersistentTrackID_Invalid)
compositionVideoTrack?.preferredTransform = CGAffineTransform(rotationAngle: .pi / 2)
let soundtrackTrack = mainComposition.addMutableTrack(withMediaType: .audio, preferredTrackID: kCMPersistentTrackID_Invalid)
var time:Double = 0.0
for (index, videoAsset) in arrayVideos.enumerated() {
let atTime = CMTime(seconds: time, preferredTimescale: 1)
try! compositionVideoTrack?.insertTimeRange(CMTimeRangeMake(start: CMTime.zero, duration: videoAsset.duration), of: videoAsset.tracks(withMediaType: .video)[0], at: atTime)
try! soundtrackTrack?.insertTimeRange(CMTimeRangeMake(start: CMTime.zero, duration: videoAsset.duration), of: videoAsset.tracks(withMediaType: .audio)[0], at: atTime)
time += videoAsset.duration.seconds
}
let outputFileURL = URL(fileURLWithPath: NSTemporaryDirectory() + "merge.mp4")
print("final URL:\(outputFileURL)")
let fileManager = FileManager()
do {
try fileManager.removeItem(at: outputFileURL)
} catch let error as NSError {
print("Error: \(error.domain)")
}
let exporter = AVAssetExportSession(asset: mainComposition, presetName: AVAssetExportPresetHighestQuality)
exporter?.outputURL = outputFileURL
exporter?.outputFileType = AVFileType.mp4
exporter?.shouldOptimizeForNetworkUse = true
exporter?.exportAsynchronously {
DispatchQueue.main.async {
completion(exporter!)
}
}
}

Answer 1

Don't use a Double to track the insertion time, this can result in gaps due to rounding errors. And don't use a preferredTimescale of 1 when converting seconds, this will effectively round everything to whole seconds (1000 would be a more common timescale for this).
Instead to track the insertion time use a CMTime initialized to kCMTimeZero, and use CMTimeAdd to advance it.
And one more thing: Video and audio tracks can have different durations, particularly when recorded. So to keep things in sync, you may want to use CMTimeRangeGetIntersection to get the common time range of audio and video in the asset, and then use result to for insertion in the composition.