Audio drift on a film

Hello! A few weeks back, I helped an editor import footage into FCPX, get it organized, sync external audio etc. Now we have found a problem whereas the synced audio slowly drifts after a couple of minutes of playback.

I did some research in here and asked Lord Google. [strike]I suspect the problem is because the video audio (scratch track) was recorded at 44khz and the external audio was recorded at 48khz[/strike]. I confirmed that the scratch track and external audio were both recorded at 48khz...so I don't think that's it.

I found some old FCP7 help articles which talked about importing the audio into compressor and exporting at the desired sample rate.

Is there a way in FCPX to correct this audio drift sync issue?

We're using FCPX 10.2.1.

Hi Clifford,

Hard to say what causes the drift.

Open the audio clip in QT and press CMD+i. This will open the clip info window. Post a screenshot of that window.

- Ronny

Though it shouldn't be the cause you might want to double check that the system audio is also set to 48k.

Another thing you might try if the file is encoded in some strange codec
is to converti the audio to aif or wav 48k via a simple QT Player or Compressor export..
then import itto fcpx as a stand alone audio clip, then do the audio sync step.

I have heard many times the 44 vs 48 Khz explanation for audio sync drift, but don't understand that. Regardless of audio sample rate the playback rate is in the same elapsed time. It's similar to 30 vs 60 fps video. If it were the audio sample rate, then 192 Khz audio would rapidly diverge from 44 Khz audio.

In this case both audio tracks were recorded at 48 Khz. Normally you don't see a problem even on fairly long takes, but depending on the timebase accuracy of the devices, some drift could creep in.

I think FCP X does a waveform match at one point of the clip and after that if the two waveforms drift, it's tough luck.

PluralEyes will do audio drift correction and continuously adjust the sync over the clip duration. That would be one workaround.

The manual way would be trying to expand or contract the external audio to match the camera audio. Sometimes this doesn't work on long takes because the drift pattern is more complex. In that case you have to cut it on scene or transition boundaries and adjust each one individually.

"I have heard many times the 44 vs 48 Khz explanation for audio sync drift, but don't understand that. Regardless of audio sample rate the playback rate is in the same elapsed time."

No, it doesn't play back at the same elapsed time. Here's how it works, and has been like this since FCP 1.0. Drop a 44.1kHZ audio clip into a 48kHz timeline. FCP(7 or X) plays every single sample, at the rate of the timeline. So the 44.1 thousand samples that made up the first second of video, get played back at 48 thousand a second. Not something you'd notice until enough seconds go by, and you'll watch the audio drift out of sync.

I've always wondered by NLE's didn't just reconfirm audio sample rates like they do video frame rates, but they don't.

Yes, this does happen. I demonstrated it in my class this week. I've experienced this first hand, over and over for over a decade and a half now.

But there can be other causes, although mismatched sample rates is the most common cause.

Thanks everyone for the responses. This type of stuff is definitely something I didn't learn in my FCPX courses.

It looks like the cause was that the DP recorded at 23.98 fps instead of 24fps, which threw it out of audio sync with the audio files. For the first minute, everything in the sync'ed clip is really good but after that the audio starts to drift. The longer the clip, the more pronounced it becomes.

When we converted the video clips to 24 fps and then did an audio sync, the audio drift was eliminated.

I'm a bit surprised by this because I didn't think external audio was recorded in "frames per second". Since it is "frames per second", I thought it was purely a video setting.

I wish there was an easy way to convert the clips within FCPX.

The DP was right to assume 23.976 fps because that's what 99% of stuff is shot in these days when you say 24p. When you're recording audio, there are settings to match whether you're doing 23.976 or true 24 fps, which is sample rate most likely. This was an important distinction when people actually shot true 24 fps in the film days (well, film is still shot at times). Soooo, in sum, I would blame the audio guy... unless there was a reason to shoot true 24 that I'm not aware of in your project.

As for doing it in FCPX, actually you can retime the audio without affecting the pitch if you wanted to, but since you know the cause, it's probably better to batch it externally and do it all at once.

Sometimes video frame rate could be the problem. Once you adjust 23.98 to 24, the program have to add 2 blank frames every 100s separately while audio cannot catch up.