Digitized Video Audio Out of Sync? Offset vs. Drift, and How to Fix Each

The two-minute test: offset or drift?

Do this before you download any tool. You need the free VLC player (Windows, Mac, Linux) and a spot in the tape where someone is talking to the camera — a birthday toast, a kid reciting something, anyone whose mouth you can see.

1. Open the file in VLC and jump to a talking moment near the start of the file.
2. Tap the J and Kkeys. Each press shifts the sound by 50 milliseconds and VLC flashes the running total on screen ("Audio delay: -300 ms"). Keep tapping until the lips and the voice agree. Write the number down.
3. Now jump to a talking moment in the last few minutes of the file and do it again.

If both numbers are about the same, you have a constant offset. That one number is your fix, and the picture never needs to be touched. If the end needs a much larger correction than the start — say 100 ms at the beginning and 3 seconds at the end — you have drift, and the audio has to be stretched to match the video's length.

Two notes on the test. First, VLC's J/K change is only for the current playback; it does not alter the file, which is exactly why it is safe to experiment with. Second, most people stop noticing a mismatch once it is under about a tenth of a second, so do not chase perfection past that — a home movie is not a dubbing studio.

Why transfers come back this way

You do not need to know the cause to fix the file, but it helps to know whether the problem will repeat on the next tape.

Causes of a constant offset

• The capture setup added a fixed delay. Cheap USB capture sticks often send the picture through one path and the sound through another, and one of them takes a little longer. The gap is the same for the whole tape.
• A DVD was converted and the converter ignored the stored delay. Many DVD recorders and services wrote the disc with a small audio delay recorded in the file itself. A converter that copies the audio without honoring that value produces an MP4 that is off by that exact amount from start to finish. This is common on files that started life as VOB. (If your file will not play at all, that is a different problem — see our won't-play guide.)

Causes of drift

• Frames were dropped during capture. When the computer could not keep up, it skipped frames of picture while the sound kept rolling. Every dropped frame pulls the picture a little ahead of the audio, and the gap grows for the rest of the file. This is the classic cause on home captures, and the reason tape restorers talk about time base correctors — a device that steadies the wobbly signal coming off an old VCR so fewer frames get dropped.
• The file's frame rate is tagged wrong. North American tape is 29.97 frames per second. If a capture tool wrote the file as 30 instead, the picture runs about 0.1 percent fast and a one-hour tape ends roughly 3.6 seconds out of sync. That number is exact enough to recognize: if your drift is about 3 to 4 seconds per hour, this is probably it.
• Sound and picture ran on separate clocks. Some capture devices take audio through the computer's line-in while video comes over USB, and two clocks never agree perfectly over an hour. Consumer MiniDV camcorders also recorded audio on its own clock, so long DV captures can slip slightly.

Drift that is smooth — the gap grows steadily — is fixable with a stretch. Drift that jumps around, in sync for ten minutes and then suddenly a second off, usually means a burst of dropped frames at one spot, and no single stretch will fix both halves. That case is covered in the re-capture section below.

Fixing a constant offset (no re-encoding)

You have your number from the VLC test. The goal is to write a new file with the audio shifted by that amount while the video is copied across bit-for-bit, so nothing about the picture changes and the job takes about as long as copying the file.

Option A: Avidemux (free, point-and-click)

Avidemux is a free editor for Windows, Mac, and Linux with an audio shift box measured in milliseconds. Open the file, leave the video output set to Copy, set the audio output to Copy as well, enter your shift in the audio settings, and save as a new file. If the saved file plays with the sound going the wrong way, flip the sign and save again — it costs a minute.

Option B: ffmpeg (free, one line)

ffmpeg is the free command-line tool most video software is built on. The trick is to open the same file twice, take the video from the first copy and the audio from the second, and delay one of them. To make the sound play later by 0.4 seconds:

ffmpeg -i tape.mp4 -itsoffset 0.4 -i tape.mp4 -map 0:v -map 1:a -c copy tape-fixed.mp4

To make the sound play earlier instead, delay the video side rather than the audio — put the offset in front of the first copy and swap the maps:

ffmpeg -itsoffset 0.4 -i tape.mp4 -i tape.mp4 -map 0:v -map 1:a -c copy tape-fixed.mp4

-c copy is the part that matters: it tells ffmpeg to move both streams across without re-encoding, so the picture quality is identical and the run takes seconds, not hours. Open the result in VLC and check a talking moment at the start and the end. If it went the wrong direction, you used the wrong one of the two lines — run the other.

Which direction is which? In VLC, a positive audio delay means the sound was made to play later. If tapping K into positive numbers fixed it, you want the first line (audio later). If tapping J into negative numbers fixed it, you want the second.

Fixing drift (stretch the audio)

With drift, the audio track is simply a different length from the video track and the two need to be made equal. You stretch or squeeze the sound by a tiny factor — well under one percent — which is far too small to change how anyone's voice sounds. The video is still copied untouched; only the audio is re-encoded.

You need two numbers: the length of the file and how far off the sound is at the very end (from the VLC test, minus whatever small offset it had at the start).

• Sound is ahead by the end (voices arrive before the mouths move) → the audio is running fast and needs to be slowed down. Factor = length ÷ (length + seconds ahead).
• Sound is behind by the end (voices arrive late) → the audio is running slow and needs to be sped up. Factor = length ÷ (length − seconds behind).

Worked example: a 60-minute file (3,600 seconds) where the sound is 3 seconds ahead at the end. Factor = 3600 ÷ 3603 = 0.99917. Then:

ffmpeg -i tape.mp4 -filter:a "atempo=0.99917" -c:v copy -c:a aac tape-fixed.mp4

Use five decimal places — over an hour, rounding to 0.999 leaves you more than half a second off. If the file also had a constant offset at the start, fix the drift first, then apply the offset fix from the previous section to the result.

Check the result at the start, the middle, and the end. If the middle is fine but one end is not, the drift was not smooth, and you are in the next section.

When the honest answer is a re-capture

A stretch fixes drift that grows evenly. It cannot fix a file where the sync jumps — in step for twenty minutes, then a second off, then back — because that pattern means chunks of picture went missing at specific moments and the audio for those moments still exists. You could cut the file at each jump and correct every piece separately, and for one precious tape that is worth an evening. For a shelf of them, it is not.

• If a service did the transfer, ask them to re-run the tape and give them a timestamp where the problem is obvious. This is a routine request and a good shop will do it. It is the strongest argument for keeping the tapes until you have watched every file through.
• If you captured it yourself, close every other program, record to a fast internal drive rather than a USB stick, and use a capture device that carries sound and picture over the same cable. Capture software usually reports dropped frames as it goes — if the counter climbs, stop and fix the setup rather than the file. The troubleshooting section of our VHS-to-USB guide covers the OBS audio offset setting for the constant-delay case.

One exception worth knowing: if the sync is bad only during stretches of snow or blue screen between recordings, it is not your problem at all — there is nothing to be in sync with. Cut those stretches out (our blank-footage guide shows how) and re-check the real footage on either side.

Fix first, then split

It is tempting to skip straight to organizing — cut the three-hour file into the birthday, the recital, the beach — and deal with the sound clip by clip. Do not. Splitting changes nothing about sync: a clip taken from minute 150 carries exactly the offset the long file had at minute 150. With drift, you would end up with forty clips that each need their own slightly different correction, instead of one file that needs one.

That includes TapeSave. It cuts clips from the file you upload and keeps the audio in the same alignment it has in that file — it does not re-time it. So the order is: run the test, fix the master with one of the lines above, confirm it at both ends in VLC, and only then upload. The corrected MP4 is accepted directly (MP4, MOV, AVI, MKV, and MPEG files all are), every clip comes out with the sound where it belongs, and the dates and labels land on footage you will actually want to share. If you are doing the splitting by hand, the same order applies — our splitting guide starts from a clean master for this reason.

And if the file has no sound at all rather than late sound, that is a separate diagnosis with a happier ending more often than you would expect — see digitized video has no sound.

Frequently Asked Questions