Recorded Tracks not synchronised

[Johannes]

There is no drift in the tracks as far as I can tell.

Yes, I guess that would work, but would probably only be accurate within 20-30ms (or more), which is no better than aligning by trial and error.

Surely there must be a better way to do this. If not, JK should be updated to have this built in. It doesn't have to be a spike, it could equally well be a cut-out for something like 1 s right after the recording has started on all the machines.

The strange thing is that JK supposedly does the alignment itself for its own mixed recordings, so there must be a way. Unfortunately, in my case the final mixes never show, so I cannot check whether there is any trace in the file.

[SteveW]

Are you saying the band wouldn't be able to hit a single stab note that accurately? I would have expected it to be much tighter than that, but have never actually tried it, so you may be right.

Or are you saying that the inaccuracy would be because of the internet latency? In which case I would have thought that doesn't matter because the audio in the track is going to be delayed by the same amount. Again maybe I'm wrong about this not having tried it :-)

[Johannes]

This is not a band, it is two classical musicians playing together. But regardless, do you know how little 20ms are? That's very little. I would like to know what the original sample synchronisation would have been (as it sounded in the stream mix), not something approximating it by that kind of amount.

No, it is not because of the internet latency, or perhaps it is also, it makes little difference. I am looking for accuracy, not guessing - accuracy in the relation of the tracks, not accuracy in the playing.

[Hans Peter Augustesen]

20 milliseconds total latency between two persons in JamKazam is like playing in the same room 7 meters apart from each other.

Try to hear your sense for latency, delay here:
> https://soundcloud.com/user851807875/pub...tly-better

"Two metronome clicks start at the same time. One of the clicks is gradually delayed compared to the other.

At the end after approx. 27 seconds, the delay is about 110 milliseconds.

Then it is all repeated, but this time in stereo with the two clicks in each channel.

At about 3 seconds after the clicks start, the delay is about 10 milliseconds - after which a normal human should be able to hear the delay."

At about 6 seconds after the clicks start, the delay is about that 20 milliseconds. It CAN be heard/noticed - but not very easy

[Yarnall]

Yes, JamKazam intends to mix the original tracks together.

And it is correct that the function worked in the past.

Until 6 months ago, or so.

But it is also correct that the function is working again now.

In the last 6 hours 9 recordings have been finished by the server mixing.

See for yourself in VIEW THE FEED

Just to get back to my original question: I am not interested in any of the mixes JamKazam makes, I will mix myself, but I would like to know how JK knows how the files sync, because looking at the files I cannot find a way  to do this other than by trial and error.

Alternatively, perhaps someone can think of a method to leave a sync mark in the files? Normally, when sound from one track bleeds into the other tracks, one could just click or clap to make an audible (and visible in the waveforms) sync mark. But because when recording in JK the tracks are completely separate in their audio content, I cannot think of an easy way to get a click or clap into all tracks simultaneously.

Or is this delay between the tracks a fixed value? (Again, I am not talking about the latency when playing together, which in our case is around 20ms. But the delay between the tracks is around 250ms, caused by the recording not starting (and stopping) at the same time. The recordings are also not the same length, the difference, however, is much larger than the delay, probably around double, but I don't think it is consistent.

Any tipps appreciated.

Maybe something as simple as starting each recording by getting a band member to count 1, 2, 3, then everyone plays a single note on their instrument, or claps their hands if they are using a vocal mic. It should be fairly easy to line up these single instrument hits and hand claps in a DAW so that at least the start of the tracks were all lined up (just like using a clapper board).

If there is significant drift in the tracks, this could be repeated before each song.

Hi Johannes, et al. - 

I have had exactly the same problems when "exporting" the master tracks to do my own mix (two musicians only, piano and bass). Our tracks are about .2 seconds out of sync (NOTE that this does VARY from song to song). And I agree with all that Johannes has said in terms of troubleshooting and analysis -- there must be a way for JamKazam to fix this automatically (and hopefully in an imminent update they will do so!), but in the meantime we must fix this manually. 

One CLUE is that when I (the bass player) export the high-res files to my Mac for mixing, it is the piano player's track that is delayed by about .2 seconds; however, the reverse is true for the piano player -- that is, when he exports the hi-res files to his Mac for mixing he sees/hears my bass track as being about .2 seconds behind. Not sure what this suggests, but it must be a clue to why/how this is happening!

What definitely DOES WORK as a solution for us (for now, until JamKazam fixes this) is essentially the same as this idea suggested earlier in this thread (a version of the "clapper" idea also referenced in this thread):

• Maybe something as simple as starting each recording by getting a band member to count 1, 2, 3, then everyone plays a single note on their instrument, or claps their hands if they are using a vocal mic. It should be fairly easy to line up these single instrument hits and hand claps in a DAW so that at least the start of the tracks were all lined up (just like using a clapper board).
  
• If there is significant drift in the tracks, this could be repeated before each song.
  

This is essentially what we do, and it works. At the beginning of each recording I give a voice-chat count-in of "1,2,3,4" after which we then we both play four clean notes in that same tempo (four taps on my bass body, four sequential high notes on the piano) in tempo with that count. NOTE that this count-in does not need to be in the same tempo as the song to be recorded; it is only being used for the sync.

Then when I export and open the files on my Mac (in a great $50 program called Sound Studio) I do the following:

1. I first open the Stream Mix in Sound Studio and listen to and look at the wave form for the count-in section.
   
2. I determine which of the four "hits" (taps/notes) is most in sync. Usually all 4 are almost perfectly in sync; but due to human error (the same as would occur in a room together, nothing to do with latency, etc.) some might have a very slight flam.
   
3. Let's say I determined that the third "hit" was the best.
   
4. I then open the hi-res version of the piano and bass tracks and merge them into one audio file (each on their own track) in Sound Studio so that I can see/hear both together (but again, with each residing on its own track within that file).
   
5. I find the count-in section on the bass track and drop in a marker right at the start of the third hit of that bass track. I then drop in a marker right at the (delayed) start of the third hit of the piano track.
   
6. I measure the distance between these two markers. Let's say that distance is .2145 seconds.
   
7. I then just delete .2145 seconds from the beginning of the piano track. This $fts that piano track back by the amount needed to cause the attack of the two most synced "hits"(on bass and piano) to align exactly.
   

This works 100% PERFECTLY every time!
It sounds complicated, but it really takes only about a minute or two to do all of the above. 
YES, it is a pain -- and JamKazam should do this automatically. But until they fix this bug this workaround will work!

[Johannes]

You know, I think I know where JK hides the info we need, and it just needs to be decoded. In the Commentary field of the Wav file I see for one file:

Start=1:1590398098:700251|JamType=LocalTrack|JamRecordingId=3c4aad44-1094-4a65-bb60-eebc7b3f5847|JamTrackId=e5a6cfda-ff46-4a90-98c3-92b614908c27|JamClientId=72c324bb-dc92-43ca-aca5-14b0895f75a5|JamLatency=521883601:0+0.0^0.0,521883601:0+0.0^0.0,715082825:0+11.0^0.2,521883601:0+0.0^0.0,521883601:0+0.0^0.0,521883601:0+0.0^0.0,521883601:0+0.0^0.0

and for the other:

Start=1:1590398098:978621|JamType=LocalTrack|JamRecordingId=3c4aad44-1094-4a65-bb60-eebc7b3f5847|JamTrackId=93d1ef0b-901a-4df2-af22-cb36682f6943|JamClientId=92f02fe2-11d8-471f-ba69-e386b833ee98|JamLatency=521883601:1+11.0^0.2,715082825:1+0.0^0.0

I would bet on it that JK leaves this info for precisely the reason to resync the files later. So if anyone knows what all of this means, let me know, it must translate somehow in the offset value of the files.

[Johannes]

I wonder whether it is in fact very simple, in that the misalignment is simply the difference between the numbers behind "Start" at teh beginning of the commentary field, with the figure after the second colon: giving the decimal value for the start of the recording. So in this case the latency would be 0.978621s - 0.700251s = 0.278370 = 278.370ms.

This seems to make sense in my case. Can anyone confirm?

[Johannes]

So this didn't work with the next test. This time the start values in the commentaries read:

Piano part:
Start=1:1590742614:731179

Violin part:
1:1590742615:20999

(Interesting that the first has one more digit).

The number would suggest a latency of around 479ms, and this certainly doesn't work. It seems to again be around 250ms, which is the value that the violin is behind, so it needs to be $fted forward by that amount to sound right. In case this makes any difference, the recording was started from the violin end.

Any ideas?

[Johannes]

I have since worked out that the latency of the piano track is 268ms or something close to that. I could work that out by using the stream mix file and finding the more or less exact beginning of the first note.

I am still hoping I can find the real value in these numbers.

This must be a simple thing to solve...

[jamsden]

I'm having the same problem. Three of us are jamming in a session. I press record, all the other jammers in the session get a notification that the session is being recorded. I press start recording and we all play. I press stop recording and go to the File Manager to wait for the high quality mix to be created. I then export the mix. The stream recording sound fine, exactly as I heard it when playing (including with dropouts and digital artifacts). The mix.ogg sounds great, but my track is quite a bit ahead of the other two.

I'm wondering if the issue is that there's no reliable time reference for syncing the high-quality tracks contributed by each jammer. When I press start recording, the recording on my machine starts - recording both the JamKazam input and my streaming output. The 0 mark in my audio files is when I pressed start recording.

At the same time, a message has to be sent to all the other performers for their recorders to start recording their high quality input and their stream. This is going to take some time for the message to be sent, for the jammer's JamKazam app to respond, and for the recorder to start recording. So the 0 time mark for the other jammer's recordings is going to be later than mine.

Is this a possible explanation? Is this a bug? Or am I doing something wrong?

I'm on MacOS 10.15.5 and am using Logic Pro X (or MainStage) with Loopback to feed audio into JamKazam