Fix audio out of sync with video.

At a glance

How to fix audio that is out of sync

Drop the video and press play. Find a moment where you can see the sound being made: a clap, a slammed door, a drum hit, or a speaker's lips closing on a "p", "b" or "m". If you hear it after you see it, the sound is late; if you hear it first, it is early. Choose which, type a first guess such as 100 ms, and play the same moment again. The preview uses the new setting straight away, so you can nudge it 10 or 50 ms at a time until the lips and the words land together. Tick "Play the original sound" to flip back and compare. When it looks right, press Save the fixed video.

Most people notice sound that comes before the picture much sooner than sound that comes after it, because that never happens in life: light arrives first, and sound from across a room is a few milliseconds behind. The broadcast guideline (ITU-R BT.1359) puts the point where viewers start to notice at about 45 ms early and 125 ms late. So when two settings both look acceptable, pick the one where the sound is slightly late rather than slightly early.

Why video and audio drift apart

A fixed offset, the same at the start and the end, usually comes from the recording chain. Screen recordings made while listening on Bluetooth headphones are a common case: Bluetooth earbuds typically add 100 to 300 ms, and some capture setups record that delay. OBS mixes with its own buffers, which is why it has a "Sync Offset" setting for each source, and capture cards, USB microphones and webcams each add their own delay. Fixing one of those is a single number.

Drift, where a video starts in sync and slides further out as it goes, is a clock problem. When sound is recorded on a separate device, such as a Zoom H1n, a Tascam, a RØDE Wireless GO or a second phone, its sample clock never runs at exactly the camera's speed. 50 parts per million is typical, which is 9 ms after three minutes and 180 ms after an hour, enough to ruin a long interview. Tick "It drifts", set how far out it is at the start and at the end, and the sound is stretched by that tiny ratio over the whole video, so every point in between lands too. A change that small cannot be heard as a change of pitch.

Lining up a separate recording automatically

If you recorded the sound on a separate device and the camera recorded its own guide track, choose "Line it up with a separate recording" and drop the recording as well. The page compares the two waveforms the way PluralEyes, DaVinci Resolve's "Auto Sync Audio" and Final Cut Pro's "Synchronize Clips" do. It uses a method called GCC-PHAT, which weighs every frequency equally rather than letting the loudest bass dominate, so a dull camera microphone across the room still matches a bright lavalier on the speaker's collar. It measures near the start and again near the end of anything over two minutes, so drift is found and corrected without you measuring anything. The check behind this page builds a camera track (muffled, with a room reflection and hiss) and a clean recording of the same three minutes of speech that started 2.5 s earlier and runs 50 ppm slow. It requires both ends to come out within 1 ms, and measures 0.09 ms at the start and 0.16 ms at the end. It also requires two unrelated recordings to be refused rather than matched.

Why the picture is not re-encoded

Editors re-encode the whole video when you change its sound, which takes minutes for a long clip and costs a little quality every time. Here the video stream is copied into the new file packet for packet. The check confirms the copied picture is byte-identical and has the same number of frames. The new sound is cut or padded to exactly the video's length, and where the move cuts into sound it fades over 5 ms, so the edit makes no click. It is written as AAC at 192 kbps, which every phone, editor and upload site plays.

What it can't fix

Phone videos are often recorded at a variable frame rate. When a player or editor handles that badly, the sync wanders back and forth instead of sliding steadily, and no single offset or straight-line drift fixes it. Converting the video to a constant frame rate in HandBrake first is the usual cure. A video where only part of the sound is out, such as a dropped stretch partway through a stream recording, needs cutting at that point in an editor: try the audio editor for the sound itself. And VLC's J and K keys, which move the sound 50 ms during playback, are a good way to measure the offset, but they change nothing in the file. This page writes the fix into the file.