In my app I use AVAssetReaderTrackOutput to extract PCM audio from a user-provided video or audio file and display it as a waveform.
Recently a user reported that the waveform is not in sync with his video, and after receiving the video I noticed that the waveform is in fact double as long as the video duration, i.e. it shows the audio in slow-motion, so to speak.
Until now I was using
CMFormatDescription.audioStreamBasicDescription.mSampleRate
which for this particular user video returns 22'050. But in this case it seems that this value is wrong... because the audio file has two audio channels with different sample rates, as returned by
CMFormatDescription.audioFormatList.map({ $0.mASBD.mSampleRate })
The first channel has a sample rate of 44'100, the second one 22'050. If I use the first sample rate, the waveform is perfectly in sync with the video.
The problem is given by the fact that the ratio between the audio data length and the sample rate multiplied by the audio duration is 8, double the ratio for the first audio file (4). In the code below this ratio is given by
Double(length) / (sampleRate * asset.duration.seconds)
When commenting out the line with the sampleRate variable definition in the code below and uncommenting the following line, the ratios for both audio files are 4, which is the expected result. I would expect audioStreamBasicDescription to return the correct sample rate, i.e. the one used by AVAssetReaderTrackOutput, which (I think) somehow merges the stereo tracks. The documentation is sparse, and in particular it’s not documented whether the lower or higher sample rate is used; in this case, it seems like the higher one is used, but audioStreamBasicDescription for some reason returns the lower one.
Does anybody know why this is the case or how I should extract the sample rate of the produced PCM audio data? Should I always take the higher one?
I created FB19620455.
let openPanel = NSOpenPanel()
openPanel.allowedContentTypes = [.audiovisualContent]
openPanel.runModal()
let url = openPanel.urls[0]
let asset = AVURLAsset(url: url)
let assetTrack = asset.tracks(withMediaType: .audio)[0]
let assetReader = try! AVAssetReader(asset: asset)
let readerOutput = AVAssetReaderTrackOutput(track: assetTrack, outputSettings: [AVFormatIDKey: Int(kAudioFormatLinearPCM), AVLinearPCMBitDepthKey: 16, AVLinearPCMIsBigEndianKey: false, AVLinearPCMIsFloatKey: false, AVLinearPCMIsNonInterleaved: false])
readerOutput.alwaysCopiesSampleData = false
assetReader.add(readerOutput)
let formatDescriptions = assetTrack.formatDescriptions as! [CMFormatDescription]
let sampleRate = formatDescriptions[0].audioStreamBasicDescription!.mSampleRate
//let sampleRate = formatDescriptions[0].audioFormatList.map({ $0.mASBD.mSampleRate }).max()!
print(formatDescriptions[0].audioStreamBasicDescription!.mSampleRate)
print(formatDescriptions[0].audioFormatList.map({ $0.mASBD.mSampleRate }))
if !assetReader.startReading() {
preconditionFailure()
}
var length = 0
while assetReader.status == .reading {
guard let sampleBuffer = readerOutput.copyNextSampleBuffer(), let blockBuffer = sampleBuffer.dataBuffer else {
break
}
length += blockBuffer.dataLength
}
print(Double(length) / (sampleRate * asset.duration.seconds))
0
1
132