I'm recording our church's sermons using a standard consumer-grade camcorder (Handycam HDR-CX380). Depending on mixer circumstances (our church moves around a lot), I'll either run a line from our mixer to the camera's microphone in, or use the camera's internal mic.
Unfortunately the audio quality isn't so hot. Lately I've been researching some options for improving the quality going into the camcorder's mic, but I'm not sure those will pay off. So what I do is split the audio from the .m2ts
file using ffmpeg
, import it into Audacity, and line it up with another file that we record on a PC (also using Audacity, but separately). I used to do this by manually subtracting timestamps, but since discovering Audacity's Time Shift tool I decided to start using that for greater precision.
But what I'm finding is that when I use ffmpeg
to extract the audio from the camcorder's audio track, there's a very small difference in the recording rate, where the camcorder's audio track is shorter than the PC's audio track by a constant rate. I haven't been able to nail down what it is precisely, but my best estimate is .20-.25 seconds per 50 minutes -- so the camera's recordings are about 99.99% the length of the PC's. It doesn't sound like much, but it causes the lip-syncing to be off about 6-8 frames.
My suspicion is either that the clock in the camcorder is bad, or that the clock in our (admittedly rather cheap-looking) USB audio interface is bad. (The error rate appears to be the same on multiple computers.) I know that I could test it using a clapboard & stopwatch, but I estimate it would take about 8 hours of dead audio/video before the rate difference got to the point that I could reliably determine the problem by looking at the waveforms and timestamps. (And that's a lot of stuff to set up for eight hours.)
So, has anyone experienced similar problems before, that could offer advice on likely causes and remedies?