Sound sync problem

Question

I have a bizarre sound synchronization problem which I am really puzzled with.

I recently recorded an interview with the video camera capturing the sound of the interviewed and an external sound capture for the interviewer.

When editing it with Adobe Premiere CS6 I started by synchronizing the start of the interview since both recordings started a few seconds apart. So far so good.

Then, one of the recordings was lagging and at the end of the recording (lasting around 25 minutes) where I found a delay of more than one second on one of the sound tracks. So I found a sync point and used the rate stretch tool to adjust one of the tracks to match the other and since beginning and ending were synchronized (I checked) all was fine... Or at least so I thought.

When I started editing further I realised that towards the middle of the recordings there was further de-synchronization. It is like the lagging and speed-up of sound is non linear; like an accordion...

Has anyone experienced this before? What causes it? How can it be corrected other than cutting everything in smaller chunks and adjusting them one by one?

Thanks in advance for any help.

PS: I had posted this same question on Video Production beta but decided to repost it here since it relates more to sound than to video.

Answer 1

Do expect drift between non-timecode-locked equipment. This is common. One way to get around it is to feed the sound from your recorder back into the camera, monitor via the camera's audio/headphones out, then use the sound from camera in your edit. Lots of cables, but fewer posts on forums later if you don't run across other complications.

Sync drift can also happen e.g. on guerrilla shoots where camera only shoots at 24p but the picture editor then reinterprets the footage as 25p, so the sound is off by 1 frame every 1 second (expect this with kit imported across US/EU borders).

Modern stretch tools (for both audio and video) can do a good job, but I wouldn't trust a computer with footage that long. I find an edit pass gives me some fatigue but overall fewer headaches.

You can also directly hack a different sample rate into a WAV file's header and then resample on import (much like pitch-shift, only we're talking fractions of a percent). I've done this with a line-in recording through audio interface vs file from recorder, which needed sample rates of 48000 and 48003 to sync perfectly (and some math to determine the drift). Soundhack does this on OSX.