Free · runs in your browser
When the sound starts in step and slips further out as the video plays, line it up near the start and near the end; the sound is stretched evenly between them.
Audio drift is sound that starts in sync and gradually slides further out: fine for the first minute, a little off by the tenth, plainly wrong by the end. A single offset can’t fix it, because the error changes over time. Instead you line the sound up at two moments, one near the start and one near the end, and the tool stretches or squeezes the whole sound track evenly so both points, and everything between, line up.
Open the video above and the Sound drifts over time section is already open. Line the sound up near the start and press Set point A, then go near the end, line it up again and press Set point B. Turn on Correct drift, play a few places to check, and press Export video. If both ends have a clap or another sharp, visible sound, Measure drift automatically sets both points for you.
Play a clear moment near the beginning: speech, a clap or a door. Nudge the sound with Earlier and Later ([ and ]) until it matches, then press Set point A.
Click near the end on the timeline and nudge again until that moment matches too. Press Set point B. The panel shows the drift in milliseconds per minute and in parts per million.
Switch on Correct drift. The badge on the preview says “drift on”, and the sound now follows the line between your two points. Play the middle of the video to check it.
Press Export video (Ctrl/⌘+Enter). The picture is copied as it is and the sound is rebuilt at the corrected speed.
Check two places far apart. If the sound is out by the same amount at the start and the end, it is a fixed offset and one number fixes it (see fix audio sync). If the gap is different, it is drift. Drift is almost always steady: a recording that is 20 ms out after one minute will be about 200 ms out after ten, because one clock was running a fraction of a percent faster than the other.
| Clock difference | After 10 minutes | After 1 hour |
|---|---|---|
| 10 parts per million | 6 ms | 36 ms |
| 50 parts per million | 30 ms | 180 ms |
| 100 parts per million | 60 ms | 360 ms |
| 0.1 % (1,000 ppm) | 600 ms | 3.6 s |
A camera and a separate recorder each count time with their own crystal. They are close, but never identical, so a long take recorded on both slowly separates. The same applies to a phone voice memo laid under camera footage. If you are joining two such files, Sync separate recordings measures the drift between them automatically.
Phones and screen recorders often save video at a variable frame rate, speeding up and slowing down to save power. Some editors read such a file as if every frame were the same length, so the picture runs slightly fast or slow and drifts against the sound. Footage that drifts in one editor but plays fine in a browser usually has this cause.
Sound recorded at 44.1 kHz but labelled or treated as 48 kHz plays about 9 % too fast, which is far too much to miss. Smaller mismatches come from devices whose actual rate is a little off their stated one.
A recorder that drops frames under load (a busy computer, a slow card) loses picture time without losing sound time. That kind of drift jumps rather than creeps, and a straight-line correction only partly helps; re-record if you can.
Measure drift automatically looks for sharp sounds and the picture changes that go with them in the first and last minutes. It needs a video of at least a minute and clear moments at both ends, and it tells you when it couldn’t find a match.
Usually because the sound and the picture were timed by clocks running at slightly different speeds, or because the video has a variable frame rate that a player or editor reads as constant. The gap then grows steadily with time.
No. An offset that fixes the start makes the end worse, and the reverse. Drift needs the sound stretched, which is what two sync points do.
There is no fixed limit; drift correction is most precise on long recordings, because the two points are further apart. Very long 4K files are easier on a computer than a phone.
Not noticeably. Real drift is a fraction of a percent, and a pitch change that small is far below what anyone can hear.
That points to a variable frame rate. Exporting the corrected file here gives the editor a file whose sound matches its picture timing as browsers play it.
No. Only the sound is rebuilt; the picture is copied into the new file unchanged.