Why Your Audio Slowly Walks Away From Your Video
You hit record and everything is perfect. Lips match words, drumsticks match drum hits. Then twenty minutes in, something feels off. By the end of the hour, the singer’s mouth is moving noticeably before you hear the voice. Nothing broke. Nothing was bumped. The sound just wandered off.
This is called sync drift, and its cause is the thing everyone gets wrong: two devices set to the same number are not necessarily running at the same speed.
Two ways of counting time
Video and audio each measure time by counting something.
Video counts frames, the still pictures shown every second (commonly 24, 25, or 30). Audio counts samples, tiny measurements of the sound wave taken tens of thousands of times a second. Video work usually uses 48,000 per second (48 kHz). Music CDs use 44,100 (44.1 kHz).
Each device counts using its own internal clock, a tiny vibrating crystal. And no two crystals are perfect twins.
The slow wander
Say your camera and your audio recorder are both set to 48 kHz. On paper, identical. But one clock might run a hair faster than the other, by an amount far too small to hear or see. Over a few seconds, nothing happens. Over an hour, the difference adds up to several frames, and your eyes are very good at spotting that.
It’s like two wristwatches that both say 3:00 today. Come back in a month and they’ll disagree, and neither is “broken.”
The loud version: a true mismatch
Sometimes the two sides aren’t merely off by a hair. They’re counting at different speeds on purpose, and the software doesn’t realize it.
Get Mtju’s stories in your inbox
Join Medium for free to get updates from this writer.
Record at 44.1 kHz but play back as if it were 48 kHz, and the player rushes through the sound about 9% too fast. The audio finishes early and the voices sound squeaky. This one is easy to catch, because it’s obviously wrong right away.
The famous version: 29.97 vs. 30
If you read our SMPTE article, you’ve already met this one. Much video runs at about 29.97 frames per second, not a clean 30. Treat the two as the same, and the gap is only 0.1%. Over an hour, that’s 3.6 seconds, which is a canyon. It’s the same quirk that gave us drop-frame timecode.
A sneaky culprit: variable frame rate
Phones and screen recorders often use variable frame rate, changing how many frames they capture depending on how much is happening on screen. That’s clever for saving space, but editing software expects steady timing. The mismatch can show up as drift, or as sync that goes wrong partway through. Your editor may not mention it, and you only find out when the lips stop lining up.
How to prevent it
- Match your settings everywhere. Use the same sample rate and frame rate in your recorder, your editing project, and your export. For video, 48 kHz is the usual choice.
- Convert variable to constant. If your footage came from a phone or screen recorder, convert it to a constant frame rate before editing.
- Share one clock when you can. Professional setups feed a single reference clock to every device, so none of them has to trust its own crystal. Timecode (see the SMPTE article) is the address telling you where you are. The reference clock is what stops devices from wandering in the first place.
- Clap at the start and the end. One loud clap when you press record and another when you stop gives you two sync points. If they don’t line up, you know exactly how far things drifted.
How to fix it afterward
If drift has already happened, most editors can time-stretch the audio, squeezing or stretching it by a tiny fraction of a percent so the ends line up with the video. Done carefully, the pitch doesn’t change and nobody can hear the adjustment. Your clap marks tell you how much to stretch.
So, to sum up
Sync drift happens because every device keeps time with its own tiny clock, and no two agree perfectly. A small mismatch, repeated thousands of times, becomes a visible gap. The fix is to make sure everyone counts the same way, ideally off the same clock, and to keep an eye on the usual suspects: mismatched settings, 29.97 vs. 30, and variable frame rates.
Two clocks that agree today won’t necessarily agree in an hour. That’s all drift is.
