Fix audio delay and lip-sync in a video
When the voices in a video arrive before or after the lips move, the sound track is out of sync. This audio sync tool moves the sound of an MP4, MOV, MKV or WebM video earlier or later by exact milliseconds, lets you watch the result before saving, and writes a new file — usually without re-encoding anything, so the picture and sound keep their full quality.
Out-of-sync video is common: phones and screen recorders that drop frames, cameras recorded with a separate microphone, Bluetooth monitoring delay during recording, or an edit exported with the wrong settings. A constant offset is fixed here in a minute. To find it, type a value and watch, or let the tool measure it from a clap and a flash in the picture.
How to use the Audio Sync / Delay Fixer
- Drop a video on the box, or press Try a generated test clip (its sound is 240 ms late).
- Play the video and adjust the Sound offset: use + when you hear words before the lips move, − when the sound comes late. The preview plays the shifted sound at once.
- Or measure it: press Detect flash and clap, or pause on the frame where the hands meet, press Mark this frame and tap the clap in the waveform.
- Choose Lossless (copies every packet) or Re-encode sound, keep length.
- Press the save button. The first time it downloads the converter (32 MB); then watch the result and download it.
Audio Sync / Delay Fixer features
Millisecond offsets
Type any shift from −10 to +10 seconds, nudge it by 10 or 100 ms, or drag the slider.
Live preview
The player plays the same file’s sound shifted by your offset, so you can judge lip-sync before saving.
Clap-and-flash detection
Finds the frame that suddenly brightens and the sharpest sound near it, and sets the offset from the difference.
Manual marks
Mark the frame where the clap happens and tap the clap in the waveform; the tap snaps to the sound’s exact start.
Lossless saving
With FFmpeg’s -itsoffset, packets are copied, not re-encoded: quality is untouched and saving takes seconds.
Private
The video is read and rewritten in your browser.
Which way to shift the sound
People notice sound that comes early much sooner than sound that comes late. The broadcast guideline ITU-R BT.1359 puts the detectability thresholds at about 45 ms early and 125 ms late, so aim for a few frames at most.
| What you notice | Offset to use |
|---|---|
| You hear words or a clap before you see it | Positive (+): the sound plays later |
| You see lips move or hands meet before you hear it | Negative (−): the sound plays earlier |
| Sync is right at the start but drifts further off by the end | Not a constant offset — the sound and picture run at different speeds; an offset alone cannot fix it |
How the two saving methods work
FFmpeg (as ffmpeg.wasm, GPL-2.0-or-later, run unmodified in your browser) opens the file twice. In lossless mode it applies -itsoffset to the second copy and takes the picture from one and the sound from the other with -c copy, so every packet is copied byte for byte. To delay the sound, the sound track starts later; to bring the sound earlier, the picture track starts later instead and the first frame is held. The file may become slightly longer.
In re-encode mode the picture is still copied, but the sound is shifted to the exact sample with FFmpeg’s adelay or atrim filter, padded or trimmed to the video’s length and encoded again as AAC (Opus for WebM). Use it when a player ignores track start times or the file must keep its length.
Tips
- Clap once in front of the camera at the start of every recording; it makes syncing later precise.
- Detection is accurate to about half a frame: ±20 ms at 25 fps, ±8 ms at 60 fps. Fine-tune by ear.
- To replace the sound with a separately recorded track, use Add Audio to Video; to cut the ends, use Trim Video.
Audio Sync / Delay Fixer FAQ
How do I fix audio delay in a video online?
Open the video here, play it and change the offset until the lips match, or detect it from a clap. Then press the save button and download the fixed video. Nothing is uploaded.
Is the video re-encoded?
Not in Lossless mode: the picture and sound packets are copied unchanged. Re-encode mode copies the picture but encodes the sound again.
Why does saving download 32 MB?
Saving uses FFmpeg compiled to WebAssembly. It is downloaded once from a public CDN and your browser caches it for later visits.
My sync drifts over time. Can this fix it?
No. A drift means the sound and picture run at slightly different speeds (for example 48 kHz sound labelled 44.1 kHz). This tool shifts the whole sound by one fixed amount.
Which formats work?
MP4, MOV, M4V, MKV and WebM. The saved file keeps the original container and codecs.
What if my clip has no flash?
Use the manual marks: pause on the frame where the hands meet, press Mark this frame, then tap the clap in the waveform.