Tutorials Video

Fixing Sync Drift Between Audio and Video

By Infinite Creation · Intermediate · ~14 min

A 1960s mid-century comic book style editorial illustration showing a video editor and a director reviewing the cut together, with a camera on a tripod visible in the frame.

Overview

Sync drift that's consistent throughout a clip (audio and video off by the same fixed offset from start to finish) has a different cause and fix than drift that gradually worsens over time, distinguishing between these two patterns is the essential first diagnostic step before attempting any fix.

What You Need

Whatever gear produced the problem. This is a diagnostic guide, not a gear-buying tutorial

Steps

First, identify whether the offset is fixed or progressive

Check the sync at the beginning and end of the same clip: if the offset is the same size throughout, it's a fixed offset (usually caused by a start-time mismatch between separate audio and video recorders). If the offset grows larger toward the end of a longer clip, it's progressive drift (usually caused by a clock rate mismatch between two separate recording devices). These have different causes and different fixes.

Fixed offset: recording start-time mismatch

A consistent fixed offset most commonly happens when audio and video were started at slightly different moments on two separate devices (a common issue with external audio recorders not synced to the camera). The fix is finding a sync reference point (a clap, a visible sound cue) and manually aligning the clips at that single point, which then holds correctly since the offset doesn't change over time.

Progressive drift: clock rate mismatch between devices

Drift that worsens over a longer recording is typically caused by two separate devices' internal clocks running at very slightly different actual rates (even when both are nominally set to record at, say, 48kHz or 29.97fps), small manufacturing tolerances accumulate into an increasingly noticeable offset over time, which a single fixed alignment point can't correct since the drift keeps growing after that point.

Fixing progressive drift, and preventing it going forward

For an already-recorded clip with progressive drift, some NLEs offer audio time-stretching tools that can gradually correct accumulating drift rather than a single fixed shift. For future recordings, using a single device to capture both audio and video, or using a genlock/timecode sync system between separate devices, prevents the clock-rate-mismatch problem from occurring at all.

Tested: how fast drift grows, and two ways to take it out

We tested progressive drift and its fix on 2026-09-24 with FFmpeg 8.1.2. We made a 20 minute track with a short clap at 5 seconds, 5 minutes, 10 minutes and 20 minutes, then made copies as a second recorder would if its clock ran slightly fast while it still labelled the file 48 kHz. We lined each copy up at the first clap, as step 2 describes, and measured how far each later clap had moved, to the sample. Then we worked out the recorder's real rate from the first and last clap and corrected the whole file two ways in FFmpeg.

Second recorder's clock	5 min after the clap	10 min	20 min	Left at 20 min after asetrate
Half a sample a second fast (48,000.5 Hz)	3.1 ms	6.2 ms	12.5 ms	12.5 ms
One sample a second fast (48,001 Hz)	6.2 ms	12.5 ms	25.0 ms	0.0 ms
0.1 percent fast (48,048 Hz)	300.0 ms	600.0 ms	1,200.0 ms	0.0 ms

The drift grew in exact proportion to time. A clock one sample a second fast was 25.0 ms out after 20 minutes, under a frame at 25 fps. A clock 0.1 percent fast, the ratio between 30 and 29.97 fps, was 300.0 ms out after 5 minutes and 1,200.0 ms after 20, a full frame at 25 fps within 40 seconds. So a mismatched frame rate setting shows within a minute, while a clock error of a sample or two a second may not show until a long take ends. Lining up at one clap fixed neither.

Two claps are enough to find the real rate: the distance between them on the recording divided by the distance on the reference, times 48,000. With it, asetrate tells FFmpeg the file's true rate and aresample converts it back to 48 kHz. That removed the drift to within a sample when the rate was a whole number. asetrate only takes whole numbers, so the recorder half a sample a second fast was left 12.5 ms out. This method changes pitch as well as timing, by 1.73 cents at 0.1 percent, under a fiftieth of a semitone.

atempo takes any ratio and removed the drift in every case, but it did two other things. It moved every clap 19.4 ms to 21.3 ms earlier. Even at a ratio of exactly 1 it dropped 1,024 samples, 21.3 ms, from the start of the file, so the clip has to be lined up again after the fix. And it stretched each clap from 4.9 ms to up to 7.9 ms, because it rebuilds the sound from overlapping pieces. Listen to sharp sounds such as drums after using it.

The fix for a recorder that ran one sample a second fast: ffmpeg -i recorder.wav -af asetrate=48001,aresample=48000 fixed.wav

Validated. Measured 2026-09-24 with FFmpeg 8.1.2 on macOS 26.6.2, Apple silicon. The drift was made by resampling, so it is perfectly steady. A real recorder's clock may not run at one steady rate, and two claps only measure its average over the take. We did not test any editing program's time-stretch tool.

Pro Tips

Always check sync at both the start and end of a clip, not just the beginning. A clip that syncs perfectly at the start can still have progressive drift you won't notice until later in the timeline.
Use a clap or audible sync cue at the start of every multi-device recording. It's the simplest fix for a fixed-offset sync problem and takes seconds to set up.
If you regularly record with separate audio and video devices, consider a timecode sync system to prevent progressive clock drift rather than fixing it after the fact.

Knowledge Base

What You'll Learn

Audio-video sync drift has two distinct patterns, fixed offset and progressive drift, with different causes (start-time mismatch versus clock rate mismatch) and different fixes, and correctly identifying which pattern you have is the necessary first diagnostic step.

Quick Diagnostic Checklist

Is the sync offset the same size at the start and end of the clip, or does it grow?
Were audio and video recorded on the same device, or separate devices?
Is there a clap or audible sync cue available at the start of the recording to align to?
Does your NLE offer audio time-stretching tools for correcting progressive drift?

When to Escalate

If drift is severe and no sync cue exists, consider whether a timecode or genlock sync system is needed for future recordings, since correcting severe undocumented drift after the fact can be difficult or impossible to do cleanly.

Where This Fits

Sync drift starts at capture, and Video Production Fundamentals explains how sound and picture choices meet in the edit.

FAQ

Q: How do I know if my sync problem is a fixed offset or progressive drift?
A: Check the sync at both the beginning and end of the same clip, if the offset is the same size in both places, it's a fixed offset from a start-time mismatch. If it's larger at the end than the beginning, it's progressive drift from a clock rate mismatch between devices.

Q: How can I prevent audio-video sync drift in future recordings?
A: Record audio and video on the same device when possible, use a clap or audible sync cue at the start of every take for easy manual alignment, and consider a timecode or genlock sync system between separate devices if you frequently experience progressive drift.