AVFoundation route audio between two non-system input and ouputs - swift

Question

I've been trying to route audio from a virtual Soundflower device to another hardware speaker. The Soundflower virtual device is my system output. I want my AVEAudioEngine to take Soundflower input and output to the hardware speaker.
However having researched it seems AVAudioEngine only support RIO devices. I've looked AudioKit and Output Splitter example however I was getting crackling and unsatisfactory results. My bones of my code is as follows
static func set(device: String, isInput: Bool, toUnit unit: AudioUnit) -> Int {
let devs = (isInput ? EZAudioDevice.inputDevices() : EZAudioDevice.outputDevices()) as! [EZAudioDevice]
let mic = devs.first(where: { $0.name == device})!
var inputID = mic.deviceID // replace with actual, dynamic value
AudioUnitSetProperty(unit, kAudioOutputUnitProperty_CurrentDevice,
kAudioUnitScope_Global, 0, &inputID, UInt32(MemoryLayout<AudioDeviceID>.size))
return Int(inputID)
}
let outputRenderCallback: AURenderCallback = {
(inRefCon: UnsafeMutableRawPointer,
ioActionFlags: UnsafeMutablePointer<AudioUnitRenderActionFlags>,
inTimeStamp: UnsafePointer<AudioTimeStamp>,
inBusNumber: UInt32,
inNumberFrames: UInt32,
ioData: UnsafeMutablePointer<AudioBufferList>?) -> OSStatus in
// Get Refs
let buffer = UnsafeMutableAudioBufferListPointer(ioData)
let engine = Unmanaged<Engine>.fromOpaque(inRefCon).takeUnretainedValue()
// If Engine hasn't saved any data yet just output silence
if (engine.latestSampleTime == nil) {
//makeBufferSilent(buffer!)
return noErr
}
// Read the latest available Sample
let sampleTime = engine.latestSampleTime
if let err = checkErr(engine.ringBuffer.fetch(ioData!, framesToRead: inNumberFrames, startRead: sampleTime!).rawValue) {
//makeBufferSilent(buffer!)
return err
}
return noErr
}
private let trailEngine: AVAudioEngine
private let subEngine: AVAudioEngine
init() {
subEngine = AVAudioEngine()
let inputUnit = subEngine.inputNode.audioUnit!
print(Engine.set(device: "Soundflower (2ch)", isInput: true, toUnit: inputUnit))
trailEngine = AVAudioEngine()
let outputUnit = trailEngine.outputNode.audioUnit!
print(Engine.set(device: "Boom 3", isInput: false, toUnit: outputUnit))
subEngine.inputNode.installTap(onBus: 0, bufferSize: 2048, format: nil) { [weak self] (buffer, time) in
guard let self = self else { return }
let sampleTime = time.sampleTime
self.latestSampleTime = sampleTime
// Write to RingBuffer
if let _ = checkErr(self.ringBuffer.store(buffer.audioBufferList, framesToWrite: 2048, startWrite: sampleTime).rawValue) {
//makeBufferSilent(UnsafeMutableAudioBufferListPointer(buffer.mutableAudioBufferList))
}
}
var renderCallbackStruct = AURenderCallbackStruct(
inputProc: outputRenderCallback,
inputProcRefCon: UnsafeMutableRawPointer(Unmanaged<Engine>.passUnretained(self).toOpaque())
)
if let _ = checkErr(
AudioUnitSetProperty(
trailEngine.outputNode.audioUnit!,
kAudioUnitProperty_SetRenderCallback,
kAudioUnitScope_Global,
0,
&renderCallbackStruct,
UInt32(MemoryLayout<AURenderCallbackStruct>.size)
)
) {
return
}
subEngine.prepare()
trailEngine.prepare()
ringBuffer = RingBuffer<Float>(numberOfChannels: 2, capacityFrames: UInt32(4800 * 20))
do {
try self.subEngine.start()
} catch {
print("Error starting the input engine: \(error)")
}
DispatchQueue.main.asyncAfter(deadline: .now() + 0.01) {
do {
try self.trailEngine.start()
} catch {
print("Error starting the output engine: \(error)")
}
}
}
For reference the RingBuffer implementation is at:
https://github.com/vgorloff/CARingBuffer
and the AudioKit example
https://github.com/AudioKit/OutputSplitter/tree/master/OutputSplitter
I was using AudioKit 4 (however the example only uses AudioKit's device wrappers). The result of this code is super crackly audio through the speakers which suggests the signal is getting completely mangled in the transfer between the two engines. I am not too worried about latency between the two engines.