Why is the new AKSequencer not producing any sound? - audiokit

Question

I have a fairly complex app that has been working with the AKAppleSequencer up until now, but due to some strange behavior and bugs that pop up now and then with that sequencer, I've been hoping to move to the newer AKSequencer. Unfortunately, the new sequencer doesn't seem to be represented in the Playgrounds or much documentation, so I have been doing some guesswork. I have everything wired up in a way that seems to make sense (to me) and, as I mentioned, was working fine with AKAppleSequencer, but with AKSequencer it runs but no output is produced.
The structure of my code is broken out into multiple pieces so the node graph gets built up in disparate locations, so I'll have to show it here in chunks, with irrelevant lines deleted.
// This happens during setup
mainMixer = AKMixer()
mainMixer.volume = volume
AudioKit.output = mainMixer
// In later code, the sequencer is constructed
sequencer = AKSequencer()
sequencer!.tempo = tempo
// After the sequencer is created, I create various nodes and tracks, like this
let trackNode = trackDefinition.createNode()
let track = sequencer.addTrack(for: trackNode)
track >>> mainMixer
There's a line up there where I'm calling "createNode()" on a thing called trackDefinition. I don't think the details of that class are relevant here, but here's an example of the body of that method's code. It's pretty straightforward.
func createNode() -> AKNode {
let pad = AKMIDISampler()
do {
try pad.loadSoundFont(partConfiguration.settings["soundFontName"]!,
preset: Int(partConfiguration.settings["preset"]!)!,
bank: Int(partConfiguration.settings["bank"]!)!)
} catch {
print("Error while loading Sound Font in PadTrackDefinition: \(error)")
}
return pad
}
That code seems to be working fine. I just wanted to illustrate that I'm creating an AKMIDISampler node, loading a soundfont, and then using that node to create a track in the AKSequencer. Then I attach the track to the main mixer for output.
I used AudioKit.printConnections() to get some confirmation, and here's what that looks like.
(1]AUMultiChannelMixer <2 ch, 44100 Hz, Float32, non-inter> -> (0]AudioDeviceOutput) bus: 0
(2]Local AKSequencerTrack <2 ch, 44100 Hz, Float32, non-inter> -> (1]AUMultiChannelMixer) bus: 0
Pretty simple... Track >>> Mixer >>> Output
Doesn't make any sound when playing.
I also tried it this way:
(0]AUSampler <2 ch, 44100 Hz, Float32, non-inter> -> (2]AUMultiChannelMixer) bus: 0
(2]AUMultiChannelMixer <2 ch, 44100 Hz, Float32, non-inter> -> (1]AudioDeviceOutput) bus: 0
So that's AKMIDISampler >>> Mixer >>> Output (and the sampler was used to create a track).
That also doesn't make any sound.
I also saw this answer to a similar question on StackOverflow, so I tried that approach. That gave me this connection graph:
(0]AUMultiChannelMixer <2 ch, 44100 Hz, Float32, non-inter> -> (1]AudioDeviceOutput) bus: 0
(2]Local AKSequencerTrack <2 ch, 44100 Hz, Float32, non-inter> -> (0]AUMultiChannelMixer) bus: 0
(3]AUSampler <2 ch, 44100 Hz, Float32, non-inter> -> (0]AUMultiChannelMixer) bus: 1
That would be [AKMIDISampler, Track] >>> Mixer >>> Output.
Still...no sound.
What am I doing wrong here? Is there some more specific way that the new sequencer tracks have to be connected into the signal graph that I'm not understanding?
UPDATE: Weird/fun/interesting addendum, if I add this code immediately after the node construction code, it produces the expected note, so I know that at least the audio engine itself is hooked up:
let midiNode = trackNode as! AKMIDISampler
try! midiNode.play(noteNumber: 60,
velocity: MIDIVelocity(127),
channel: MIDIChannel(8))

Answer 1

I figured this out, and wanted to post the answer here for future developers who may run into confusion around this, and also for the core AudioKit team to see, so they can understand what might not be obvious from the API.
The root of the problem here was that the AKSequencer is not a drop-in replacement for the AKAppleSequencer, even though the APIs for the two are extremely similar.
One thing to point out: I have confirmed that it is in fact necessary to add both the track itself and the track's target node to the signal chain in order to get sound output. So from my examples above, you need this one:
[AKMIDISampler, Track] >>> Mixer >>> Output
This is sort of weird and confusing, because it's not at all obvious where I would be expected to put effects nodes in between those. I haven't played with that yet, but it seems very strange to have these nodes both be siblings in the signal chain. I would think it would look like this:
Track >>> AKMIDISampler >>> Mixer >>> Output
That makes more sense to me. Oh well.
Anyway, I mentioned that there were some other factors that were the root of the problem. The key difference was that with the AKAppleSequencer, the track lengths could start out at 0 and then grow as you added additional notes to them. This is the approach I was using, as I was starting with empty tracks and then populating them procedurally.
With the new AKSequencer, it doesn't appear to work that way. The length starts out as 4.0, not 0, and it does not grow automatically as you add notes to the tracks. I had to manually calculate the length required to fit my notes, and then set that length using track.length = desiredLength. The good news is, the AKSequencer is able to understand to use the length of the track, so you can set it on just the tracks, and not the sequencer itself if you prefer.
Another notable difference is the behavior of stop() on the sequencer. On the AKAppleSequencer, invoking stop() also stops the playback of all the notes. On the new AKSequencer, the same method will leave notes playing. You need to do a loop over the tracks like this:
sequencer.stop()
for track in sequencer.tracks {
track.stopPlayingNotes()
}
I know the AKSequencer is brand new, so some things like this are to be expected. I still have hope that it is going to be better in the long run than the AKAppleSequencer.
I hope this explanation will help out somebody like me who got stuck switching to the new sequencer!