Audio timeline programming glitch

[Geoffrey Cox]

I have noticed what looks like a programming error in Premiere, niche but serious.

The problem occurs when you use the audio time units scale / display to fine tune the position of audio files, then try and move them with the video.

Scenario:

3 mono audio tracks, two a stereo ambient pair, the other, a radio mic on the 'talent'
1 video track (with muted linked audio, used for synch only)

After lining up the three linked audio clips with the video (via a slate tone), due to phasing issues it is sometimes necessary to move the radio mic clip a few milliseconds (I use the milliseconds scale rather than samples). This works fine to eliminate the problem, and my whole capture setup was designed with this possibility being needed in mind.

I then re-link the three audio clips and group them with the video clip so they can all be easily moved as one unit without losing any synch.

EXCEPT when you do move a group, the fine timing adjustment to the radio mic clip is lost! The software inexplicably re-aligns it with the other audio clips even though the audio time units scale is still enabled and the clips are linked and grouped. Worse, if you 'undo' it, the realignment sticks! It doesn't undo it. Only loading a previous version gets you back to where you were. All of this occurs if you cut and past as well as click and drag

This does NOT occur if you use the roll tool or close a gap in the timeline by selecting it and hitting backspace.

Given that I have many edits and many 'Groups' that contain this small time adjustment of the radio mic clips, this is a very major issue for me.

The only work around I have found is to de-group, move the linked audio first, then align the video afterwards. In other words if you move or cut and paste the linked audio only, the automatic software re-alignment error does not occur. Having to move audio and video independently is a real pain though and also prone to error, - any grouped video clip that you forgot to ungroup will cause the error, and remember, Undo doesn't work!

I get this is a bit niche but surely having to deal with audio phasing / combing issues like this is far from uncommon? I did find a couple of people online complaining about this very problem but with no solutions. These complaints go back many years so this is not new.

Anyone else noticed this, got any suggestions? I have tried all manner of solutions but only moving the audio and video separately works (as long as audio time unites is enabled, if it is not, moving these files independently will cause the re-alignment, but that is at least logical, as the timeline only has the much coarser frames scale to align with)

Thanks for reading.

[Christopher Young]

Over the years, mainly with mirrorless or DSLR jobs, I have come across many situations where un-synced audio has had to be realigned by a few samples or microseconds at the head of the clips.

In addition to this problem, I find on long audio takes sync drift can occur over time, this drifting loss of sync can be pretty severe at the end of long clips even if the sync was perfect at the start of the clips. To eliminate these problems during the edit, I'll fix the issues before I start the edit by using Pluraleyes. Saves countless hours. Pluraleyes will do this fine micro-alignment of the unsynced audio clips with the video audio and will also apply drift correction to maintain sync over time by extending or shrinking the new unsynced clips it creates. On the timeline, you can then group and move the groups around without losing the fine micro syncing that Puraleyes created when making its new audio clips.

But if I don't have that software to do the job and have to do a manual sync, like it sounds like you are doing, I'll do the following. I'll line up and sync everything along a dedicated "sync" timeline or timelines, and make sure the ends of each audio clip aligns with the ends of its appropriate video clip. Once every clip is correctly micro synced, I then do a batch render and export all the finally realigned audio clips from each of their dedicated "sync" tracks. In other words, creating a series of new .WAV files for each track. If it's only one track causing issues, then you only need to create one set of new audio files for that track. Batch rendering a bunch of WAV file is very quick.

Next, I'll mute the track or tracks with the finally synced audio clips. Then I create a new audio track or tracks and bring in these new WAV audio files. They will align perfectly with their corresponding video clips and once grouped with their respective video clips can be moved around without any slippage of sync. Should you lose sync for any reason, it's easily fixed by just aligning the start your video clip with the start of its new WAV partner. This method of course will not fix audio sync drift issues on long takes.

Premiere can do all of the above syncing and audio drift correction, working in conjunction with Pluraleyes. Or without Pluraleyes do batch audio exports for reimport.

Unfortunately, unless you use one of the above methods, I think you will be battling uphill with the problems you outline. Unless of course someone else can chip in with a better solution... hopefully!

Chris Young

Sync all your video and audio at once! PluralEyes + Adobe Premiere - YouTube

[Geoffrey Cox]

Thanks Christopher, good thoughts.

I did consider 'bouncing' the audio to a single stereo file and re-synching with the video but the problem is that would remove any possibility of further mixing between the original.tracks. Without going into detail, this would be prohibitive, and suffice to say, as I have lots of short 'vignettes' to eventually bring together in a single timeline, I am bound to need to do more fine tuning of the audio mix right up to the completion of the edit.

I came across this work around which looks better than the one I am currently using. Yet to try it.

https://community.adobe.com/t5/premi...idi-p/13603588

[Paul R Johnson]

I have to own up and say I didn't believe you, and I created a duplicate of what you have - the camera audio, plus three other tracks - I used a project where there was a clap - nice and easy to align. So I changed the timeline to the audio smaller divisions, and aligned the tracks. I grouped them, and they stayed where I had put them. Aha! I thought. Then I moved the whole thing a few frames and they split apart!. This was on a mac, so I tried on Windows with the same result. What is the point of the extra zooming in on the audio tracks if the thing reverts to to clip lengths. Sorry I doubted you - doesn;t solvce the problem though. I've never noticed as I usually sit with the other tracks up to half a frame late from the video without noticing - it only annoys me when I get the nasty phasing when both audio tracks have to be heard .

[Geoffrey Cox]

I don't blame you for being sceptical Paul. I was convinced I was simply doing something wrong but after lots of tests I had to come to the conclusion I wasn't and it is a fault.

Either that or the programmers assumed you would never want to move material around once the initial timing adjustments had been made. But that's silly.

I don't understand how it works under the hood but clearly there are two different timescales in operation at the same time - frames only for the video but down to samples for the audio. I would have thought that in effect these are not really different scales as a single frame can be divided into samples or milliseconds too, but when you move linked a/v material it simply defaults to frames come what may and any sample based adjustments in the audio is chucked in the bin, even with the audio time units scale enabled. And as you say, what is the point of that? There is none. It's a mistake.

[Christopher Young]

Originally Posted by Geoffrey Cox

I came across this work around which looks better than the one I am currently using. Yet to try it.

Yes, it does seem like a work around that would work. Shame we have to resort to workarounds. That's why I went to Pluraleyes years back. It's saved me countless days over the last twenty years. Eliminates all the issues you've mentioned.

One thing I forgot to mention is if you want to work with videos of different frame rates. Like some footage at 24p, some at 50p and some at let's say 29.97 this is something Pluraleyes can't handle correctly. If I have a sync job with mixed frame rates I use Syncaila which does a very similar job to Pluraleyes, but it will handle mixed frame rates correctly while syncing.

Chris Young

Syncaila Review - Perfect for syncing audio and video! - YouTube

[Geoffrey Cox]

Thanks Christopher. I watched the Pluraleyes video and it looks good. How it might deal with multiple audio tracks for the same interviews isn't clear though and this is essentially the root of the problem I have since one of those tracks (the vocal mic)) needs fine timing adjustment to get rid of the out-of-phase issue with the other audio tracks. So the problem isn't so much about synching the audio with the video but getting the timing relationship between the audio tracks correct for good sound. I am struggling to see that Pluraleyes would understand this specific problem and deal with it but I might be wrong.

You might ask why do I need three audio tracks for an interview and that is a good question but there are very real reasons for it, suffice to say they aren't so much interviews as performances that need stereo ambience in addition to a mono voice track and these need to be recorded at the same time. This means the voice is also picked up by the stereo pair (as well as the ambience), and this is important to what I am trying to achieve soundwise.

[Patrick Tracy]

I work in Vegas Pro so some specifics don't apply, but this general workflow might. When I have a more involved audio mix, I typically do it as a separate project, render a stereo track and import that and sync it up to the video. If I need to revise the audio, I just go back to that project and re-render it. The new audio file will update in the main video project with all the edits intact as long as the start point in the audio project stays the same.

One could get even fancier and nest the audio mix project inside the video project as a track, but I rarely use that feature and I don't know if other video editors can do that.

[Geoffrey Cox]

For info, the send to Adobe Audition doesn't really work. It could have worked if I had done it to the original edits right at the start of the process (and I had some success trying it on one file) but since I first realigned them, I have done lots of detailed editing and trying the Audition trick on these edits, messed the sound up in a whole new way.

[Christopher Young]

Originally Posted by Patrick Tracy

I work in Vegas Pro so some specifics don't apply, but this general workflow might. When I have a more involved audio mix, I typically do it as a separate project, render a stereo track and import that and sync it up to the video. If I need to revise the audio, I just go back to that project and re-render it. The new audio file will update in the main video project with all the edits intact as long as the start point in the audio project stays the same.

One could get even fancier and nest the audio mix project inside the video project as a track, but I rarely use that feature and I don't know if other video editors can do that.

But that is some of the magic in Vegas Pro. For audio, nothing came close to it for years. Back in between 2000-2010 doing weekly TV shows we used to finish the shows in Avid and or Discreet edit* and then finish all the audio in Vegas in conjunction with Sound Forge. Where you can open SF in the Vegas timeline and non-destructively make very fine audio edits in SF from the Vegas timeline. And very rarely did you ever have sync or phase issues to deal with, as everything stayed as you had adjusted and wanted it. But there again Vegas' background was that it was programmed by people from Sonic Foundry who were DAW knowledgeable programmers who understood audio.

Chris Young

[Christopher Young]

Originally Posted by Geoffrey Cox

Thanks Christopher. I watched the Pluraleyes video and it looks good. How it might deal with multiple audio tracks for the same interviews isn't clear though and this is essentially the root of the problem I have since one of those tracks (the vocal mic)) needs fine timing adjustment to get rid of the out-of-phase issue with the other audio tracks. So the problem isn't so much about synching the audio with the video but getting the timing relationship between the audio tracks correct for good sound. I am struggling to see that Pluraleyes would understand this specific problem and deal with it but I might be wrong.

You might ask why do I need three audio tracks for an interview and that is a good question but there are very real reasons for it, suffice to say they aren't so much interviews as performances that need stereo ambience in addition to a mono voice track and these need to be recorded at the same time. This means the voice is also picked up by the stereo pair (as well as the ambience), and this is important to what I am trying to achieve soundwise.

That's what I was trying to get at. You can easily fix the issues you outline. Why you develop out of phase issues is due to a number of things. One is different sample rates between sources. Even though ostensibly they may all be defined as 44KHz, 48KHz or whatever, the crystal oscillators in the cameras and recorders can often be clocking at fractionally different rates. Like fraction of a Hz difference. So while all tracks may be "synched" unless you have software that will adjust drift and maintain phase between all your audio sources, you will always run into these sorts of issues. What programs like Pluraleyes do is to use an audio track off one of the videos, the one closest to a true 48KHz, and synch and phase adjust all other video and audio tracks to that truest 48KHz track. Just between un-genlocked different brand cameras alone, we have rncountered a number of frames difference in synch, which affects phase as synch is lost. Over a period of an hour a number of frames drift in synch and phase is a massive post-production problem to fix.

The question? Can Pluraleyes handle and keep in sync and maintain phase of a number of wild ambient soundtracks in conjunction with video tracks along with their associated soundtracks, regardless of the number of tracks involved or how the audio has been split across those tracks. The answer? Yes it can. That's how my business has done it in producing countless corporates and TV shows for local and overseas markets when working with non-time code or non-genlocked sources.

Unless you have time code or genlocked sources, where one master crystal has the other cameras and recorders slaved to it, I don't know how you would maintain 100% sync and phase. I would love to know if there is another way because in countless years of editing I have not come across any.

Finally, what a lot of people don't know is that with broadcast cameras, frame audio is re-clocked and locked to the start of vision scan line one of each frame. This guarantees that the audio and video from that camera can never cause any issues with genlocked cameras and recorders. Sadly, this is NOT the case with all prosumer gear.

Give me the old way any day. Time code and genlock. Never had these sync and phase issues in post with field recorders and cameras that supported those pro standard synch locking capabilities

Chris Young

[Patrick Tracy]

I didn't see if it was mentioned, but if all the audio sources are recorded to one device, i.e. with one ADC clock, drift shouldn't be a problem. Any audio phase mismatch should be essentially static. A simple shift right or left of a track will fix it. But here it looks like those simple fixes are reversed when groups are moved. It doesn't look like a drift problem, it looks like a software design flaw, like audio track positions are quantized to some reference too coarse to keep the phase relationships stable at audio frequencies.

[Geoffrey Cox]

Thanks for your further thoughts Christopher and Patrick.

I should have been clearer that I know exactly why I am getting the out of phase problem and it is simple physics: the time delay between the person speaking and radio mic (mounted on their shirt or whatever) is slightly less than that hitting the stereo pair mounted on the camera. At a reasonable distance of a couple of metres or so this is not noticeable but as I get closer in, it is, sometimes (the angle affects it too). There are basic mathematical formulae for this (e.g. about 3ms delay per metre).

In the field I set a small time delay of about 5ms on the radio mic in the multi-track audio recorder (a Tascam mounted under the camera) which mitigates the problem but I like to move about handheld to get more dynamic footage so adjusting this in real time is impossible, plus I found monitoring it accurately also impossible as you hear too much of the direct sound outside of the headphones.

I always knew this would be an issue, hence how important it is to be able to finely adjust the radio mic clip position in the Premiere timeline when necessary (about 30% of the clips probably). By soloing this track and comparing it is very simple to remove the out of phase / combing / hollow sound. But I didn't bargain on Premiere's faulty programming.

You might ask why I work this way in the field, with the three mics: it is born out of a desire to capture real ambience in synch with the voice (like those leaves rustling in the wind behind the speaker), to get a richer voice sound than the (mono) radio mic that naturally blends with the ambience, and that I work entirely alone.

[Geoffrey Cox]

Originally Posted by Patrick Tracy

I didn't see if it was mentioned, but if all the audio sources are recorded to one device, i.e. with one ADC clock, drift shouldn't be a problem. Any audio phase mismatch should be essentially static. A simple shift right or left of a track will fix it. But here it looks like those simple fixes are reversed when groups are moved. It doesn't look like a drift problem, it looks like a software design flaw, like audio track positions are quantized to some reference too coarse to keep the phase relationships stable at audio frequencies.

Yes, exactly, it is a programming error in my view. The quantisation is always to the frame when you move any audio that is attached to video, so all audio is quantised to the frame no matter that you have previously adjusted some of the audio finer than a frame. If you move the audio on its own, the adjusted phase relationships are retained.

[Patrick Tracy]

Originally Posted by Geoffrey Cox

Yes, exactly, it is a programming error in my view. The quantisation is always to the frame when you move any audio that is attached to video, so all audio is quantised to the frame no matter that you have previously adjusted some of the audio finer than a frame. If you move the audio on its own, the adjusted phase relationships are retained.

This is why I suggested doing the audio work first in a separate project and importing a finished stereo mix. Then any phase correction between tracks would be fixed into the stereo file. You could even trim the top of the file or add a small bit to account for the quantization to whole video frames that is happening in the video project.