Audio/video sync puzzle

Question

I filmed an event with two cameras and I need to sync the second camera footage with the audio of the first. This was a live event and I couldn't record sync aids like a clapper.

I name my video files after the timestamp from the camera, and I figured I could use this to do the sync. So for instance, I have one clip with name VID-2017-04-22-11-09-26.mov and another with name VID-2017-04-22-11-16-22.mov, so I can work out that the second clip ended 6 minutes and 56 seconds after the first. (The camera only knows when you finished filming and created the file, and the naming convention puts hh-mm-ss as the last three pairs of numbers.)

So I figured that I could just note the spot on the timeline where clip 1 ends and place clip 2 so that it ends at that time plus 6:56.

But when I did this, it was totally wrong, and I can't figure out why. The only variable I can think of it that the frame rate of this footage is 59.94 while the sequence is 23.976fps. But I'm not sure if that's relevant, and if so, I still can't figure out how to place these clips.

Meanwhile, I'm stuck syncing the very hard way, by watching the clip and manually trying to find the matching footage from camera 1, and this is painfully tedious.

Update

From comments posted, I realize I wasn't clear. Both clips are from the same camera, which is camera 2. I have synced the first of these clips from camera 2 manually with camera 1, and now I need to sync the remaining clips from camera 2 with each other.

Update

The workflow was this: I set camera 1 on a tripod to film the speaker continuously (but rather boringly). I walked around with camera 2 for more dynamic angles, sometimes pausing between shots. So camera 2 has about 25 different clips, which I want to line up with camera 1, and then cut between them to make a more interesting video.

Answer 1

I record conferences all the time, so I end up getting 10's of hours of footage at a time. I have a few tricks that help with syncing:

1. Run all cams all the time. If a full day shoot gets only 3 files per cam, that's a lot less work to sync than 25 files from each cam.
2. Stop and start all cams within the same few seconds.
3. Learn to notice key waveforms. Things like applause, audience chatter, and music have unique shapes. You can also watch for "um" and long pauses.
4. There's sync tools available, like PluralEyes, that analyze the sound from each clip, then syncs them on the timeline.
5. There's time code methods that many pros use, but I've never had any success with them.

I depend on 1 through 3 almost exclusively. I've gotten very good at reading waveforms and I only stop cams for breaks longer than 10 minutes. All my clips are 30 seconds out of sync at most. It's pretty easy to sync a whole day in 15 minutes. I've used PluralEyes, but found it took too long to process the material, and it was off by about a half second most of the time, so I had to go though and move all the clips anyway. I only use it if I have tons of clips to sync (more than 15), instead of my typical four or five. PluralEyes will get you close, then you have to adjust manually.

Answer 2

If the time (clock) settings for the two cameras are different, the 'end-time' relationship is meaningless. If you still have access to the cameras, start them both and record a clap. Ingest the two short clips and note any difference in time at the clap. This difference, plus or minus, should be added to the 6:56 to get a true offset. You can trim by frames from there.

In any case it shouldn't be hard to find a matching audio event in the two clips. Any distinct sound can stand in for a clapper. Of course if your software has audio auto-alignment this becomes trivial.

You may still have drift issues since the two cameras might not be clocked exactly the same, but once you have the gross offset, little trims are easier.

Answer 3

From what I can see, your method requires two assumptions to be made (and I don't think either one can be safely made here).

The internal time of the cameras isn't synced. The two cameras you shot with probably used internal clocks that you set the first time you turned them on. They're not connected to the internet or getting updated with any sort of precision. You could manually change the time of one of them to 5:00 PM and the other to 2:30 AM and they'd operate normally. What may be 11:16:22 AM on one camera is probably more like 11:15:54 AM on the other.

Seconds aren't precise enough for syncing audio/video. If you are syncing with just using hours/min/seconds, you're not syncing as precise as if you were using frames, which in your case would be 1/24 of a second (in your sequence). Even if the internal clocks on both cameras were perfectly synced, it's not like you're starting and stopping recording at exactly 1 second, 0 frames each time.

Answer 4

Well, I feel a little dumb because the answer is that this camera actually does timestamp the video file with the starting time, NOT the ending time. So I can accomplish what I need by computing the time difference from the filenames and offsetting the start of a clip from the start of the first clip. Duh.