Topic: Tools for fixing the sync drifting recordings?

[faninor]

What software do you all use for sync problems? I use Adobe Audition for most everything, but when it comes to sync problems I end up wasting a lot of time guessing and checking on the stretch ratio, and if the drift doesn't occur in a linear fashion it is a real headache.

What I would really enjoy is software that would allow me to load two audio files, and then set anchor points in each file... so I could create a set of points in file A that correspond to another set of points in file B, although they might not line up correctly due to drift. Then after these anchor points are all set it it processes file B so that its anchor points actually do line up with the corresponding points in file A... and the synchronization for everywhere in between is interpolated intelligently. Does a free or affordable tool that does something like this exist?

If so, it would be a huge time saver for me. So much that I'm trying to convince my girlfriend that we should make such a tool if I can't find one.

[boojum]

Can your girlfriend code C++ and assembler? 

[faninor]

Can your girlfriend code C++ and assembler? 

Sure. Her degree is in computer science, and mine is in mathematics. So I'm pretty sure we can work something out if we need to. 

I was bored at work this afternoon so I came up with algorithms I would need in order to take two given sets of anchor points and then find the displacement at any given sample in the output & dither it. I still need to work out how to apply some sort of low pass filter to avoid distorting the highest frequencies. I have the general idea for it but need to come up with the details. Alternatively I can probably skip that and just use files with high sample rates, then convert the sample rate down using other software.

[stantheman1976]

What exactly are you trying to sync together?  Are you trying to mix two audio sources?  Are you replacing the audio on a cideo recording?

If you're trying to combine two recordings of the same even and they are both from digital sources you can use Sony Vegas to easily match them up.  As you know, digital devices run at a constant speed so if there is drift between the sources it will be constant.  With Vegas you can line the two sources up at the beginning of the recording either by ear or using a common visual reference like a hand clap or cymbal hit that shows up as a distinct point in the waveform.  You can literally hold down a key and stretch or shrink the track in real time.  Hit me up with a PM if you decide to try Vegas and I'll go into detail.

[faninor]

What exactly are you trying to sync together?  Are you trying to mix two audio sources?  Are you replacing the audio on a cideo recording?

If you're trying to combine two recordings of the same even and they are both from digital sources you can use Sony Vegas to easily match them up.  As you know, digital devices run at a constant speed so if there is drift between the sources it will be constant.  With Vegas you can line the two sources up at the beginning of the recording either by ear or using a common visual reference like a hand clap or cymbal hit that shows up as a distinct point in the waveform.  You can literally hold down a key and stretch or shrink the track in real time.  Hit me up with a PM if you decide to try Vegas and I'll go into detail.

I want to use it for several purposes, including mixing together transfers from analog sources where the drift is not necessarily linear. I didn't know Vegas could do that... that sounds more user-friendly than Audition but probably not quite what I'm looking for, it sounds like I'll have to split the track up a bunch and match up a lot of points if the drift isn't linear. I would prefer something that uses more reference points and interpolates what is happening between them.

[Brian Skalinder]

You could use Samplitude SE, split the WAV into multiple objects and apply stretch/shrink independently to each object (I think this is the same solution as what stan suggested with Vegas).  Not automated, but I think it will accomplish what you wish.  You might be able to do the same in Audition by using the multitrack view, and instead of splitting the WAV into multiple objects, split into multiple clips.  Each clip will have it's own track, which could get tedious.  Not sure, though, it's been a while since I've used Audition.  I don't know of a way to automate the Sam/Vegas option.

[Jammin72]

You could use Samplitude SE, split the WAV into multiple objects and apply stretch/shrink independently to each object (I think this is the same solution as what stan suggested with Vegas).  Not automated, but I think it will accomplish what you wish.  You might be able to do the same in Audition by using the multitrack view, and instead of splitting the WAV into multiple objects, split into multiple clips.  Each clip will have it's own track, which could get tedious.  Not sure, though, it's been a while since I've used Audition.  I don't know of a way to automate the Sam/Vegas option.

I was toying around with this Trying to mix the AKG C34 Source FOB from Panic NYE with the Schoeps MK41->148 Section source.  What kind of workflow would be needed to make this happen.

I'm coming up with.

Load Each set of Tracks Per Set. 
Join Said Tracks
Slide to match a beginning point.
Stretch to match an ending point (either one)
Bounce

Thing is that Visually stretching and Alignment isn't giving me the control over the points I need to make it happen.  They jump a bit.

Maybe I need to do it in smaller sections but that seems to be an incredible pain in the ass.


While it stays aligned though... 

[Brian Skalinder]

Thing is that Visually stretching and Alignment isn't giving me the control over the points I need to make it happen.  They jump a bit.

Get it close by stretching visually, then...

<1>  Set the unit of measurement to milliseconds (View | Units of Measurement | Milliseconds)

and fine-tune one of these two ways...

<2a>  Zoom in on the endpoint of the object you're timestretching.  The closer you zoom in, the finer the increments while visually timestretching.

or

<2b>  Offline Effects or Realtime Effects | Object Resampling / Timestretching and set the Time Factor Calculation's New Length manually

[Jammin72]

Thing is that Visually stretching and Alignment isn't giving me the control over the points I need to make it happen.  They jump a bit.

Get it close by stretching visually, then...

<1>  Set the unit of measurement to milliseconds (View | Units of Measurement | Milliseconds)

and fine-tune one of these two ways...

<2a>  Zoom in on the endpoint of the object you're timestretching.  The closer you zoom in, the finer the increments while visually timestretching.

or

<2b>  Offline Effects or Realtime Effects | Object Resampling / Timestretching and set the Time Factor Calculation's New Length manually

As if you needed another.... +T


Wonder if Reaper will let me do this.  It may be back to Sequoia for that type of work.

[faninor]

You could use Samplitude SE, split the WAV into multiple objects and apply stretch/shrink independently to each object (I think this is the same solution as what stan suggested with Vegas).  Not automated, but I think it will accomplish what you wish.  You might be able to do the same in Audition by using the multitrack view, and instead of splitting the WAV into multiple objects, split into multiple clips.  Each clip will have it's own track, which could get tedious.  Not sure, though, it's been a while since I've used Audition.  I don't know of a way to automate the Sam/Vegas option.

That would work, but I may still want to create my own solution since I want to try to cut down on tedious activities.

I have a question for those of you who have done this with either program: does the wav file get processed and actually change when you strech/shrink it, or only when it is rendered / mixed down to an output file?

[Brian Skalinder]

That would work, but I may still want to create my own solution since I want to try to cut down on tedious activities.

I have a question for those of you who have done this with either program: does the wav file get processed and actually change when you strech/shrink it, or only when it is rendered / mixed down to an output file?

Yeah, cutting down on the tedious tasks would be of great benefit for the kind of thing you're doing, no doubt.  Regarding when the changes take place, it depends on the editor.  Non-destructive editors will only apply the changes when rendering, while destructive editors will apply changes on-the-fly.  Samplitude SE offers both options.  Not sure about Vegas or the other apps.

[stantheman1976]

Vegas is non-destructive.  It doesn't change anything until you render your file.

[stantheman1976]

What exactly are you trying to sync together?  Are you trying to mix two audio sources?  Are you replacing the audio on a cideo recording?

If you're trying to combine two recordings of the same even and they are both from digital sources you can use Sony Vegas to easily match them up.  As you know, digital devices run at a constant speed so if there is drift between the sources it will be constant.  With Vegas you can line the two sources up at the beginning of the recording either by ear or using a common visual reference like a hand clap or cymbal hit that shows up as a distinct point in the waveform.  You can literally hold down a key and stretch or shrink the track in real time.  Hit me up with a PM if you decide to try Vegas and I'll go into detail.

I want to use it for several purposes, including mixing together transfers from analog sources where the drift is not necessarily linear. I didn't know Vegas could do that... that sounds more user-friendly than Audition but probably not quite what I'm looking for, it sounds like I'll have to split the track up a bunch and match up a lot of points if the drift isn't linear. I would prefer something that uses more reference points and interpolates what is happening between them.

If you can come up with something then good luck but I don't think it will happen.  I read the latest issue of DV magazine and the people who are digitizing and releasing the first seasons of SNL talked about their process of syncing up better audio sources and how long it took.  If these guys who are using professional level equipment can't just use a formula to calculate the difference then I don't think we could do any better.

[joemango]

Here's a link to the Gathering of the Vibes Tapers' board...  I'd rather link than cut and paste so you can see the whole discussion.  In my experience the stretch function in Sound Forge doesn't have the resolution needed to make this fine of an adjustment.  There's some math involved... read on.

http://www.gatheringofthevibes.com/bored/YaBB.pl?num=1162332750

Hope it helps.

[it-goes-to-eleven]

As you know, digital devices run at a constant speed so if there is drift between the sources it will be constant.

Oh how I wish that was true....   The temperature of the devices varies during the recording. As a result the clock differences tend to vary.  Some devices (744t) use a higher quality crystal oscillator that is much less prone to drift (they have internal heaters so the crystal is always held at the same high temp).

[dklein]

Hey - been a while since I was on the board but I had this thread set to notify and it's just come back to life!  Here's a little tool that I've been using for a few years now.  It's pretty self-explanatory and will help you do the math behind the resampling required to bring things back into alignment.  I've found that once you know the relationship between two devices it's pretty consistent.  I used to do multitrack on separate devices.  Now the only time I use it is to match up separate audio and video (synch to the camera's crappy audio and your good audio will match the picture).

Here's my experience with some different approaches.
Stretching - I have not found stretching to sound very good, although Wavelab was much better than CEP/Audition.  Neither allow the precision required.  I seem to recall wavelab ignores everything after the first decimal point, or maybe didn't even use the first decimal point.
Cutting / matching - your always drifting off and then realigning so you're never really on the mark.  Weird phase artifacts are introduced.
Deleting samples - similar to above
Resampling - the way to go.  If you think about it, the original event occurred over the same duration to both machines.  If one is shorter, it's being played back faster and the pitch will be slightly up.  Resampling to an adjusted rate but then playing back at the 44.1/48k playback rate should bring the pitch back on while addressing the time drift issue. 

R8Brain is free and really good at it.  Always max out the quality settings...there shouldn't be any rush for this kind of work!  Let me know if you find the worksheet useful.

david

[joemango]

Cool... I didn't know r8brain could do fractional sample rates.  I'll have to use that next time.

[kbergend]

Hey - been a while since I was on the board but I had this thread set to notify and it's just come back to life!  Here's a little tool that I've been using for a few years now.  It's pretty self-explanatory and will help you do the math behind the resampling required to bring things back into alignment.  I've found that once you know the relationship between two devices it's pretty consistent.  I used to do multitrack on separate devices.  Now the only time I use it is to match up separate audio and video (synch to the camera's crappy audio and your good audio will match the picture).

Here's my experience with some different approaches.
Stretching - I have not found stretching to sound very good, although Wavelab was much better than CEP/Audition.  Neither allow the precision required.  I seem to recall wavelab ignores everything after the first decimal point, or maybe didn't even use the first decimal point.
Cutting / matching - your always drifting off and then realigning so you're never really on the mark.  Weird phase artifacts are introduced.
Deleting samples - similar to above
Resampling - the way to go.  If you think about it, the original event occurred over the same duration to both machines.  If one is shorter, it's being played back faster and the pitch will be slightly up.  Resampling to an adjusted rate but then playing back at the 44.1/48k playback rate should bring the pitch back on while addressing the time drift issue. 

R8Brain is free and really good at it.  Always max out the quality settings...there shouldn't be any rush for this kind of work!  Let me know if you find the worksheet useful.

david

I just want to thank you for posting this!  I had to align mic and sbd sources from two different recorders (R-44 and PMD620) last night in Audition and wasn't having much success with the stretch function.  I found this post, downloaded the free version of r8brain, followed the instructions in the spreadsheet, and it worked perfectly!  No noticeable degradation in the slightly faster soundboard resampled to 44,100.33 using Highest Quality, and it saved me hours of tedious work correcting for the drift piecemeal.

+t! (if I live to 500)

[boojum]

What a board!     

[jaz]

  Tap the "R" key in Wave Lab and it brings up the Wave Restore Tool. It only works for a small sections but a digi pop from sync issue usually fits into it's range. BUT through the global analysis of errors it seldom finds digi pops as slight as the ones that occur do to sync issues. 


Or if you got the cash a lynx L22 and you wont have sync issue's.

[kirk97132]

Hey - been a while since I was on the board but I had this thread set to notify and it's just come back to life!  Here's a little tool that I've been using for a few years now.  It's pretty self-explanatory and will help you do the math behind the resampling required to bring things back into alignment.  I've found that once you know the relationship between two devices it's pretty consistent.  I used to do multitrack on separate devices.  Now the only time I use it is to match up separate audio and video (synch to the camera's crappy audio and your good audio will match the picture).

Here's my experience with some different approaches.
Stretching - I have not found stretching to sound very good, although Wavelab was much better than CEP/Audition.  Neither allow the precision required.  I seem to recall wavelab ignores everything after the first decimal point, or maybe didn't even use the first decimal point.
Cutting / matching - your always drifting off and then realigning so you're never really on the mark.  Weird phase artifacts are introduced.
Deleting samples - similar to above
Resampling - the way to go.  If you think about it, the original event occurred over the same duration to both machines.  If one is shorter, it's being played back faster and the pitch will be slightly up.  Resampling to an adjusted rate but then playing back at the 44.1/48k playback rate should bring the pitch back on while addressing the time drift issue. 

R8Brain is free and really good at it.  Always max out the quality settings...there shouldn't be any rush for this kind of work!  Let me know if you find the worksheet useful.

david

Bumping this for dklein's great step by step process.  Thanks and I'm looking forward to using it to fix a very problematic mix of sources.  It's 3.5 hours of recording and even though it is one two of exactly same decks there is a drift issue.  At the end of the show it's pretty significant Even if I tried splitting it into tracks the size difference is audible by the end of the song.  This looks like the perfect solution to my problem.  I've tried stretching in Adobe Audition and just can't seem to get it right.  And I'm not real sure of work flow in WL or Samplitude.  Thanks again!  Kirk