MTAudioProcessingTap with kMTAudioProcessingTapCreationFlag_PostEffects not reflecting AVAudioMix volume - ios

Question

I am trying to build level metering for AVPlayer. I am doing this with an MTAudioProcessingTap that gets passed to an AVAudioMix, which in turns gets passed to the AVPlayerItem. The MTAudioProcessingTap gets created with the kMTAudioProcessingTapCreationFlag_PostEffects flag.
Technical Q&A QA1783 has the following to say about the PreEffects and PostEffects flags:
When you create a "pre-effects" audio tap using the kMTAudioProcessingTapCreationFlag_PreEffects flag, the tap will be called before any effects specified by AVAudioMixInputParameters are applied; when you create a "post-effects" tap by using the kMTAudioProcessingTapCreationFlag_PostEffects flag, the tap will be called after those effects are applied. Currently the only "effect" that AVAudioMixInputParameters supports is a linear volume ramp.
The problem:
When created with the kMTAudioProcessingTapCreationFlag_PostEffects, I would expect the that samples received by the MTAudioProcessingTap would reflect the the volume or audio ramps set on the AVAudioMixInputParameters. For example, if I set the volume to 0, I would expect to get all 0 samples. However the samples I receive seem to be totally unaffected by the volume or volume ramps.
Am I doing something wrong?
Here is a quick an dirty playground that illustrates the problem. The example sets the volume directly, but I observed the same problem when using audio ramps. Tested on both macOS and iOS:
import Foundation
import XCPlayground
import PlaygroundSupport
import AVFoundation
import Accelerate
PlaygroundPage.current.needsIndefiniteExecution = true;
let assetURL = Bundle.main.url(forResource: "sample", withExtension: "mp3")!
let asset = AVAsset(url: assetURL)
let playerItem = AVPlayerItem(asset: asset)
var audioMix = AVMutableAudioMix()
// The volume. Set to > 0 to hear something.
let kVolume: Float = 0.0
var parameterArray: [AVAudioMixInputParameters] = []
for assetTrack in asset.tracks(withMediaType: .audio) {
let parameters = AVMutableAudioMixInputParameters(track: assetTrack);
parameters.setVolume(kVolume, at: kCMTimeZero)
parameterArray.append(parameters)
// Omitting most callbacks to keep sample short:
var callbacks = MTAudioProcessingTapCallbacks(
version: kMTAudioProcessingTapCallbacksVersion_0,
clientInfo: nil,
init: nil,
finalize: nil,
prepare: nil,
unprepare: nil,
process: { (tap, numberFrames, flags, bufferListInOut, numberFramesOut, flagsOut) in
guard MTAudioProcessingTapGetSourceAudio(tap, numberFrames, bufferListInOut, flagsOut, nil, numberFramesOut) == noErr else {
preconditionFailure()
}
// Assume 32bit float format, native endian:
for i in 0..<bufferListInOut.pointee.mNumberBuffers {
let buffer = bufferListInOut.pointee.mBuffers
let stride: vDSP_Stride = vDSP_Stride(buffer.mNumberChannels)
let numElements: vDSP_Length = vDSP_Length(buffer.mDataByteSize / UInt32(MemoryLayout<Float>.stride))
for j in 0..<Int(buffer.mNumberChannels) {
// Use vDSP_maxmgv tof ind the maximum amplitude
var start = buffer.mData!.bindMemory(to: Float.self, capacity: Int(numElements))
start += Int(j * MemoryLayout<Float>.stride)
var magnitude: Float = 0
vDSP_maxmgv(start, stride, &magnitude, numElements - vDSP_Length(j))
DispatchQueue.main.async {
print("buff: \(i), chan: \(j), max: \(magnitude)")
}
}
}
}
)
var tap: Unmanaged<MTAudioProcessingTap>?
guard MTAudioProcessingTapCreate(kCFAllocatorDefault, &callbacks, kMTAudioProcessingTapCreationFlag_PostEffects, &tap) == noErr else {
preconditionFailure()
}
parameters.audioTapProcessor = tap?.takeUnretainedValue()
}
audioMix.inputParameters = parameterArray
playerItem.audioMix = audioMix
let player = AVPlayer(playerItem: playerItem)
player.rate = 1.0