automation timing destroyed by bouncing

[yks]

Good day to you!

Logic 10.7.4 on a new 16 inch M1 Macbook: I am mixing a big project with lots of volume automation for dialogue.

When I bounce, the volume automation is unpredictably WAY out of time: my dialogue is being cut off unexpectedly, and the problem seems to be worse the further in time into the mix I go (the total duration is about 37 minutes). By the time I get to 12 minutes, the dialogue is totally being mangled. I am not positive, but lots of my other automation may be affected as well -- effect parameters and panning.

I have searched this forum and seen that others have had the same terrible problem of unpredictable volume automation bounces. Please can someone tell me there is a solution that really works for non-realtime bouncing? I can bounce in realtime, obv, and that should work since playback in Logic is fine. But we have to be able to trust a non realtime bounce too!

Thanks for any hints.

 

[David Nahmani]

Under Logic Pro > Preferences > Audio > General, is Sample Accurate Automation set to everything?

[yks]

Thank you for the reply! Yes I have sample accurate selected for "Volume, Pan, Sends, Plugin Parameters".

[yks]

Sooo... I just was running off a non-realtime bounce, and dammit the automation is just as flaky. Only 10 minutes in I was already hearing dialogue being cut off.

I am wondering if I simply do a playback to another app to record the master output might work -- but now I am wondering if playing from the beginning is the problem -- maybe the automation drifts no matter what kind of playback is being done. I will try that next to see...

This is kind of disastrous.

In another recent project where the automation was being flaky in the same way, I ended up doing spot bounces-in-place for sections where the automation was being ignored/badly interpreted by Logic when doing nonrealtime bounce. That solved my problems that time but this session is bigger and is more riddled with places where the automation is bad.

 

 

 

[Atlas007]

5 hours ago, yks said:

I ended up doing spot bounces-in-place for sections where the automation was being ignored/badly interpreted by Logic when doing nonrealtime bounce. That solved my problems that time but this session is bigger and is more riddled with places where the automation is bad.

I was about to suggest bouncing in place for every tracks in the entire project…

[polanoid]

18 hours ago, yks said:

the automation drifts

Automation does not drift. That's impossible given how Logic plays back songs. Can you post an example project so we can get to the real cause of your issue?

Edited September 24, 2022 by polanoid

[David Nahmani]

18 hours ago, yks said:

I just was running off a non-realtime bounce, and dammit the automation is just as flaky. Only 10 minutes in I was already hearing dialogue being cut off.

Can you share the project so someone can give it a try? 

[yks]

Hi again and thanks for continued replies. The person who said that the timing of automation cannot drift hasn't heard how my bounced master mixes are so f*cked up. It sounds like, as time goes on, the automation becomes separated in time from the audio and so things are cut off unexpectedly. Please know that I am a long time Logic user for professional projects and I have favoured it over Pro Tools or some other DAWs for big, complex projects. I have never had problems with bouncing volume automation until now.

The realtime bounce I tried failed. I could hear the drift happening as it was bouncing and stopped the process.

I then tried running off the project by playing it back into another DAW (via Loopback) to record the 2track output. I thought that might have fixed the problems but the client is saying they still hear issues.

I will try to create a version of the project which has a reduced number of tracks, isolating a couple that I know were very problematic. If I can share a simpler version of the project which also shows the problem, I will do that.

[David Nahmani]

28 minutes ago, yks said:

The person who said that the timing of automation cannot drift hasn't heard how my bounced master mixes are so f*cked up

I believe the issue is with the word "drift". In this industry drifting is typically used to describe the progressive loss of synchronization, for example between two machines that should play in sync, and they start in time, but one of them progressively goes out of sync because it goes slightly faster than the other. And indeed, that is not possible for automation to drift in a Logic project, so wile you may experience an issue with automation being out of time, it's most likely not what we normally refer to as "drift". 

28 minutes ago, yks said:

I will try to create a version of the project which has a reduced number of tracks, isolating a couple that I know were very problematic. If I can share a simpler version of the project which also shows the problem, I will do that.

Let's do that, that is the best way for someone to experience and understand what the issue is, and avoid all sorts of miscommunication. 

[Atlas007]

What kind of automation is at stake? RBA, TBA, faders, CCs, trimmed, relative?

[Atlas007]

1 hour ago, David Nahmani said:

And indeed, that is not possible for automation to drift in a Logic project

Just out of curiosity, could you please explain how it is impossible for automation data to not sound as getting (gradually) out of sync (drift) when bounced?

(Not intending to challenge your statement here, just wishing to understand…)

[RobertAnthony]

I know what OP is talking about and it SUUUUCKS.

I had this happen and after an hour just said “screw it”, reverted to the previous day’s project file, and lost a day of work recreating stuff, vs spending an unknown amount of time trying to fix the project if it could have even been fixed.

I’m hoping OP has a manually saved previous version to fall back on?

I’ve got the “F’d up” project file that I couldn’t fix saved if that would help in any way (another file exhibiting the same symptoms, etc.). I made a quick video when the issue initially occurred, but I never did anything with it.

Edited September 25, 2022 by RobertAnthony

[Atlas007]

3 hours ago, RobertAnthony said:

I’ve got the “F’d up” project file that I couldn’t fix saved if that would help in any way (another file exhibiting the same symptoms, etc.). I made a quick video when the issue initially occurred, but I never did anything with it.

Sure do!

It couldn't hurt...

[polanoid]

13 hours ago, yks said:

The person who said that the timing of automation cannot drift hasn't heard how my bounced master mixes are so f*cked up

Then there must be another reason. Automation does *not* drift. Just put some MIDI notes next to it and you will hear that they are always synchronized. I suspect that it’s rather the audio that isn‘t played back / bounced in time. Waiting for an example project to clear that up.

[polanoid]

5 hours ago, RobertAnthony said:

project file that I couldn’t fix saved if that would help in any way

Of course it would. Please make it available for download so people on this forum can have a look.

[polanoid]

12 hours ago, Atlas007 said:

ust out of curiosity, could you please explain how it is impossible for automation data to not sound as getting (gradually) out of sync (drift) when bounced?

Automation is played back via the sequencer (in the event list you can see that it consists of Events just like the regular MIDI events) and the sequencer always plays back in the order given by the event position in the timeline. So no event can ever overtake another one, they will always be in sync with the musical grid. Hence my suspicion that it's rather the Audio File playback that gets out of sync.

[yks]

I appreciate that the volume automation should be locked in place but clearly something is happening to destroy that. I haven't had time to prepare a sample session for you, sorry. Perhaps today I will have time.

An additional major problem in this session is that I am using many "summing stacks" and when I move tracks around internally between stacks they don't necessarily pick up the new routing for the new assignment (ie they remain routed to the old track stack unless I notice and change it manually). I just spent half an hour debugging bounces which I am doing SCENE BY SHORT SCENE for this large radio play project, in order to try and avoid the volume automation problems, because it seems that the problem occurs when trying to bounce long duration mixes. I am having to track down missing sounds from scenes because of the bad routing. BAD LOGIC PRO. I am not used to saying that but Logic is really failing me on this new M1 Macbook.

[David Nahmani]

Just now, yks said:

An additional major problem in this session is that I am using many "summing stacks" and when I move tracks around internally between stacks they don't necessarily pick up the new routing for the new assignment

In order for us to efficiently help troubleshoot issues, it's better if we focus on only one issue at a time in one topic, so that each separate issue has its own topic, with its own descriptive title, tags, and its own thread of steps to help troubleshoot. It's better for you, for us, but also for everyone else in the future who's searching the web for one of these issues. 

So in this topic let's keep the focus on the automation timing issue, but feel free to start a new topic for the summing stack routing issue so that we can help you with that separately. 

[yks]

Dear Polanoid, I am looking for this other reason, but none seems to exist. The software option which should solve this problem (checking the box in Audio Options for sample accurate bouncing) does nothing for me, unfortunately.

In my session, the audio regions remain in place, just as you are suggesting to do with a MIDI event (I am not needing any MIDI elements in this session). But the volume automation is very definitely off, in relation to those audio regions.

 

5 hours ago, polanoid said:

Then there must be another reason. Automation does *not* drift. Just put some MIDI notes next to it and you will hear that they are always synchronized. I suspect that it’s rather the audio that isn‘t played back / bounced in time. Waiting for an example project to clear that up.

 

[yks]

17 hours ago, Atlas007 said:

What kind of automation is at stake? RBA, TBA, faders, CCs, trimmed, relative?

The automation is volume + FX parameters -- reverb dry/wet mix balance changes for instance. Nothing unusual or particularly complex.

In some of the radio play scenes I am working on, there is dialogue between two characters which was recorded without good isolation, so I am manually "noise gating" with volume automation (one of the characters is on the phone so there is a radical tonal difference, thus any bleed is to be avoided). The automation is cutting very close to the dialogue so any drift is very audible. I am hearing LOTS OF DRIFT when doing a full-length bounce of the master mix, and also when trying to bounce all the short scenes individually, by selecting them all and trying to do a BIP for all. In the latter case Logic started at the beginning of the session and gave me back a single long WAV (not what I expected since I had rendered many individual scenes, which were their own unique regions, and unique regions should I think result in separate bounces). The timing of the automation was off starting around minute 12.

My latest strategy, which seems to be working, is to manually bounce each scene (they are each defined as a track stack, so I put an empty MIDI region on that lane to BIP it). Logic seems to be OK with short duration bounces and all the automation stays in place. Or maybe it is already starting to drift, but the amount is negligible.

[yks]

Caveat: I did a NRT test bounce of the full length of the reduced session linked below and despite that there were major problems in the bounce of the full session, there were NO VOLUME AUTOMATION PROBLEMS FOR THESE SCENES. I can only guess that having removed the other files from the project, it changed the complexity of the bouncing situation. I am interested to see if you can duplicate the volume automation timing problems but am guessing this test session is not going to help you hear that, and I can't share the full session because it's not my IP.

I am supplying a link here to a WeTransfer of the session which includes all the tracks and automation but only the audio for dialogue which was affected in two scenes, both of which involve one person heard on a telephone. The scenes begin at 22:09 and 29:28.

I am using plenty of third-party commercial plugins and also my own impulse responses with Space Designer, but you should be able to open this demo session and see dialogue for the two scenes way down at tracks 97-98 and 128-129 (screen shots attached). These tracks are part of scenes defined as track stacks.

WeTransfer link (69MB): https://we.tl/t-NPGYf7wsE5

Thanks for any comments!

 

[polanoid]

2 hours ago, yks said:

there were NO VOLUME AUTOMATION PROBLEMS FOR THESE SCENES.

So. Now we're getting somewhere. We now have a version of the project in which the issue happens, and one in which it doesn't. First thing would be to list all the things you removed from the session. Audio Files? Plug-ins? Whole channel strips?

[RobertAnthony]

For Context:
This is the other thread I commented on when I had this automation issue happen. I feel these reported issues are either the same issue or very closely related, but wanted to keep the threads separate just for continuity.

My Comments for This Thread:

11 hours ago, yks said:

I did a NRT test bounce of the full length of the reduced session linked below and despite that there were major problems in the bounce of the full session, there were NO VOLUME AUTOMATION PROBLEMS FOR THESE SCENES. I can only guess that having removed the other files from the project, it changed the complexity of the bouncing situation.

Confusion loves company I guess lol.

I spent the day messing with my "messed up" project file and I experienced the same dang thing lol (though I'll admit idk what 'NRT" stands for? Non-Real Time?).

The full project would have the automation issues, but no matter how I tried to strip the project down (delete tracks then save, or make a new project and import the offending tracks), they seemed to play as expected.

I've added the stripped down project just to add it, though I don't feel it has much value given stripping down the project seems to change something and the issue goes away. I don't really wanna upload the entire, unfiltered project file.

I was, however, able to document the issue fairly well in my opinion. I'll add the video here, but if it is better suited on the above linked Thread, I can move it there (Again, trying to not cross-pollinate the topics):

Note: As of posting this, the video is still being rendered into the HD formats. It's currently only viewable at 320p but that should change within an hr or so.

 

Project_Stripped_Rob_Anthony.zip

Edited September 26, 2022 by RobertAnthony
Added video quality note.

[polanoid]

3 hours ago, RobertAnthony said:

I was, however, able to document the issue fairly well in my opinion.

What I don't understand is - how did you make the project play back correctly when you play it for comparison?

Also I'm pretty sure this is a different issue than the OP's because it happens in realtime playback as well, not only when bouncing?

[polanoid]

3 hours ago, RobertAnthony said:

The full project would have the automation issues, but no matter how I tried to strip the project down (delete tracks then save, or make a new project and import the offending tracks), they seemed to play as expected.

You could try removing the tracks one by one and check after each track?

[RobertAnthony]

50 minutes ago, polanoid said:

What I don't understand is - how did you make the project play back correctly when you play it for comparison?

It’s two different project files. I manually save project versions, so I’ve got yesterday’s file, last weeks file, so on.

[RobertAnthony]

50 minutes ago, polanoid said:

You could try removing the tracks one by one and check after each track?

I guess I could. I’ll do that if I find a free chunk of time this week as that’s potentially a lot of deleting, saving, trying lol.

[polanoid]

10 minutes ago, RobertAnthony said:

It’s two different project files. I manually save project versions, so I’ve got yesterday’s file, last weeks file, so on.

Can you tell the difference between the two files then?

[RobertAnthony]

Maybe my video wasn’t as clear as I thought lol.

The difference is on the newer, “working“ project file, I deleted all the automation on the shown Track Stack, and all contained tracks, then just re-created it all. I spent maybe 90 minutes trying to fix the issue then, again, just said “Screw it, it’ll be faster to just delete then recreate the automation. I just want to move on”.

Assuming Save As makes an exact carbon copy of the file, that should be the only difference.

If you read the comment I left on the other Thread, idk what changed or what happened, but from my POV, I walked away from an open project, came back 30 mins later, hit play, and it was playing out of time/sync/whatever.

The steps I then took, after spending time figuring out what the problem was and doing the troubleshooting steps listed on the other thread are:

1. Did Save As and saved a copy of the project file
2. Opened the new project file
3. Deleted all the automation on the Track Stack and all contained tracks for the entire project duration
4. Saved, closed, opened for good measure
5. Annoyingly recreated all the automation like OP in this thread is trying to avoid doing, understandably so.

The way I process OPs initial message, I feel it’s the same issue. I would have used the word “shifted” instead of “drifting” to describe the automation, as drifting implies movement, like “at minute 1 the automation is in time, at minute 5 it’s slightly off, at minute 10 it’s even more off” so on.

Edited September 26, 2022 by RobertAnthony
Typos

[polanoid]

8 minutes ago, RobertAnthony said:

drifting implies movement, like “at minute 1 the automation is in time, at minute 5 it’s slightly off, at minute 10 it’s even more off” so on.

Exactly, as I already wrote, *drifting* would be absolutely impossible.