Topic: Matrixing questions

[guitard]

So we pick one and stretch or shrink the other, slightly altering its playback rate, to match.

We could adjust both, shrinking one and stretching the other to "meet in the middle" but that's twice as much work and doesn't make any significant difference.  Practically, when deciding which to alter, it might be best to choose whatever you deem to be the "secondary source", the one that is contributing less to the end result, thereby eliminating the chance of introducing artifacts into the primary source

On a semi-related note, I almost always record both audio and video; and rarely use the camera's audio.  I just sync my separate audio, but I don't sync it to the camera audio.  Instead, I sync it to the video itself.  And depending on how far the video is shot from the sound source, that can be as much as a 1/3 of a second.

[morst]

On a semi-related note, I almost always record both audio and video; and rarely use the camera's audio.  I just sync my separate audio, but I don't sync it to the camera audio.  Instead, I sync it to the video itself.  And depending on how far the video is shot from the sound source, that can be as much as a 1/3 of a second.

I sync to the camera sound, because then it comes in the way our brains would hear it from the camera position.
I can see why your technique is a good idea, though, especially for multi camera productions. Might as well sync to the closest of the batch.

[lpmaskman]

When syncing audios together, or audio to video, the delay might occurs, so you need to timestretch or timeshrink one of the source.
I mostly record audio and video together (as above said) the mostly out of sync if I place them together, that's why I timestretch it to match tha camera's audio length.

[guitard]

When syncing audios together, or audio to video, the delay might occurs, so you need to timestretch or timeshrink one of the source.
I mostly record audio and video together (as above said) the mostly out of sync if I place them together, that's why I timestretch it to match tha camera's audio length.

With modern video cameras and audio recorders, I usually don't have to change the length of the audio (either stretching or shrinking).  If there is any drift, it's likely not going to be more than one camera frame (1/30th of a second for 30 fps video) for a 90 minute video.  Even then I might not bother with it because it is virtually impossible for your eyes and ears to notice a difference on playback when stretching audio for a 90 minute video by the equivalent of one camera frame.  That's stretching a 90 minute video by 1/30 of a second.  Or in terms of camera frames, that's stretching an audio from the equivalent of 161,999 frames to 162,000 frames.

So, the only real syncing for me is getting the audio to match the action on the stage.  Where you can run into issues with that is if the band uses big screens - that are often behind by a half second or so.  But that's a whole other discussion about syncing.

[nulldogmas]

With modern video cameras and audio recorders, I usually don't have to change the length of the audio (either stretching or shrinking).

I just did my first successful show recording where I synced audio to video, and that's exactly what I found — I had to do lots of nudging of the audio file to get it to line up properly, but the sync didn't drift notably over the course of the show:

https://www.youtube.com/watch?v=qBQpqTy_nBU

With two audio files, even a tiny fraction of a second offset will be audible as an echo. With audio plus video, your brain should make the adjustment so long as they're close.

[morst]

With modern video cameras and audio recorders, I usually don't have to change the length of the audio (either stretching or shrinking).  If there is any drift, it's likely not going to be more than one camera frame (1/30th of a second for 30 fps video) for a 90 minute video.  Even then I might not bother with it because it is virtually impossible for your eyes and ears to notice a difference on playback when stretching audio for a 90 minute video by the equivalent of one camera frame.  That's stretching a 90 minute video by 1/30 of a second.  Or in terms of camera frames, that's stretching an audio from the equivalent of 161,999 frames to 162,000 frames.

Sounds like you got lucky on clocks. 1600 samples over 90 minutes is really tight. (1/30 sec at 48000 is only 1600 samples)

[guitard]

With modern video cameras and audio recorders, I usually don't have to change the length of the audio (either stretching or shrinking).  If there is any drift, it's likely not going to be more than one camera frame (1/30th of a second for 30 fps video) for a 90 minute video.  Even then I might not bother with it because it is virtually impossible for your eyes and ears to notice a difference on playback when stretching audio for a 90 minute video by the equivalent of one camera frame.  That's stretching a 90 minute video by 1/30 of a second.  Or in terms of camera frames, that's stretching an audio from the equivalent of 161,999 frames to 162,000 frames.

Sounds like you got lucky on clocks. 1600 samples over 90 minutes is really tight. (1/30 sec at 48000 is only 1600 samples)

I don't think it's really about luck anymore.  It's more a matter of how far consumer-level recording technology has come.

But what the hell do I know??

[yltfan]

I don't think anyone mentioned it yet, but for me, it depends a ton on the SBD mix - if it's a typical wonky small room mix (heavy on the vocals, light on the guitars) then you gotta lean more on the AUD sources.

[morst]

I don't think anyone mentioned it yet, but for me, it depends a ton on the SBD mix - if it's a typical wonky small room mix (heavy on the vocals, light on the guitars) then you gotta lean more on the AUD sources.

Sure, the smallest rooms will have board mixes with nothing but vocals.


All you can do to get a solid sound there, is add enough of the board feed to clean/sweeten/tighten the vocal sound and be done with it!

[jcable77]

With modern video cameras and audio recorders, I usually don't have to change the length of the audio (either stretching or shrinking).  If there is any drift, it's likely not going to be more than one camera frame (1/30th of a second for 30 fps video) for a 90 minute video.  Even then I might not bother with it because it is virtually impossible for your eyes and ears to notice a difference on playback when stretching audio for a 90 minute video by the equivalent of one camera frame.  That's stretching a 90 minute video by 1/30 of a second.  Or in terms of camera frames, that's stretching an audio from the equivalent of 161,999 frames to 162,000 frames.

Sounds like you got lucky on clocks. 1600 samples over 90 minutes is really tight. (1/30 sec at 48000 is only 1600 samples)

I don't think it's really about luck anymore.  It's more a matter of how far consumer-level recording technology has come.

But what the hell do I know??

With the different cameras I have, and different recorders and such, Ive never had to time stretch any audio to sync with video. I always use 24/48 so Im not sure if that makes a difference, but what I thought was getting lucky for a while in the beginning has become pretty common place.

[BlueSky71]

I am gonna jump on the bandwagon here. I commonly use a F6, and a F3 both in 32/24. I never have an issue. But, I time align in the middle of recordings or sets. So if I did it would be worse at the beginning or end? I dunno. Not hearing it, also not looking for it.

[Gutbucket]

[..]I time align in the middle of recordings or sets. So if I did it would be worse at the beginning or end?[snip]

Here's the answer (which I realize you weren't really looking for) to the above question.

Aligning in the middle of the recording without stretching or shrinking one to sync perfectly with the other will "split the difference" and cause any mis-synchronization to manifest equally at both start and end, but with only half the offset that would otherwise occur at the end if the files were aligned at the start instead (or vice versa).  Because doing so effectively halves the alignment error, this makes for a good strategy to avoid stretching/shrinking, as long as it successfully keeps the mis-sync at either end from being perceivable or otherwise messing with something.

Specifically, the mis-sync will manifest as the slightly shorter file (the one recorded with the slightly faster clock reference) trailing the other maximally at the start, that error being gradually reduced until the sync point, then increasing again to the same amount at the end, but with the shorter file pulling ahead of the other in time after the sync-point.

With the different cameras I have, and different recorders and such, Ive never had to time stretch any audio to sync with video. I always use 24/48 so Im not sure if that makes a difference, but what I thought was getting lucky for a while in the beginning has become pretty common place.

Sample rate and bit-depth won't make a difference. It's probably more modern recording gear incorporating more accurate clocks, good luck, and recordings that may seem long but are not long enough for the problem to become noticeable. If a recording were to run for many many hours you would be more likely to notice it. 

As Guitard noted, syncing audio and video is less problematic by two orders of magnitude in comparison to the accuracy needed for mixing two closely-correlated audio sources without phase problems occurring when the sources are mixed. If the two audio sources are not highly correlated its more like one order of magnitude before the delay will begin to become apparent.  Yet recorders which include time-code have historically tended to use more accurate clocks to begin with, partly as a way of allowing for longer free-running of time-code without problems.

[morst]

Gutbucket nails it regarding midpoint alignment. I can think of nothing to add on that subject.
 

[BlueSky71]

Yes, that's exactly how I pictured it put into words. I always thought that without changing the length of either track, aligning in the middle of the recording (roughly) would halve any potential sync issue(roughly).. Maybe I read that in a tutorial?

[morst]

Yes, that's exactly how I pictured it put into words. I always thought that without changing the length of either track, aligning in the middle of the recording (roughly) would halve any potential sync issue(roughly).. Maybe I read that in a tutorial?

we've discussed it on this website before