Audio Out of Sync in a Screen Recording: Causes and Fixes

Voice landing after the click? What really causes out-of-sync audio in screen recordings — Bluetooth, device switching, dropped frames — and how to fix it.

You click a button, say “and here it saves,” and play the recording back — except your voice arrives a beat after the click. Or the webcam bubble is nodding to a sentence it has not said yet. Or the first thirty seconds are fine and the last two minutes slide further and further apart.

All three look like the same problem and none of them have the same cause. Out-of-sync audio is one of the few recording faults where guessing at a fix wastes the most time, because a constant offset and a growing drift come from completely different places. This guide sorts them out in the order you can actually test them.

First: Which Kind of Desync Do You Have?

Watch the playback twice — once at the start, once at the end — and compare how far off the audio is in each spot.

What you observeWhat it points to
Same offset at the start and the endA fixed startup delay between tracks
Small at the start, larger at the endDropped frames, or an output device that is delaying what you hear
Only the webcam is late; voice matches the screenWebcam startup lag
The voice is right, but the video inside your recording is offThe source app dropped frames while playing
Your voice sounds doubled, with a short slapMic and system audio both captured the same sound

That table is the whole diagnosis. Everything below is just the fix for each row.

How Recorded Keeps Tracks Aligned

It helps to know what is actually on disk, because it explains which problems the app can solve for you and which ones it cannot.

Recorded does not mix everything into one file while you record. Each recording gets a folder — under ~/Movies/Recorded on macOS, %USERPROFILE%\Videos\Recorded on Windows — containing the pieces separately:

  • recording.mp4 — the screen capture
  • audio.wav — system audio, when you enabled it
  • mic_audio.wav — your microphone, when you enabled it
  • webcam.mp4 — the camera, when you enabled it
  • track_sync.json — the timing that ties them together

That last file is the important one. Every track starts at a slightly different moment in real time: a microphone stream takes a few dozen milliseconds to open, a camera can take noticeably longer, and system audio only starts producing packets once something is actually playing. Recorded timestamps each one against the screen capture and writes the difference down in milliseconds — positive when the track started after the screen, negative when it started before.

The editor reads those offsets and shifts each track by that amount during preview, and passes the same numbers to the exporter so the MP4 is muxed with the same alignment you previewed. This is why the common case — “the mic opened 80 ms late” — never reaches you as a problem. It is also why the editor has no manual audio-offset slider: the offset is measured, not guessed.

What that mechanism cannot fix is a track that was recorded wrong. If the sound arriving at your ears was already delayed, or if the picture lost frames mid-capture, the timestamps are honest and the recording is still off. Those are the causes worth working through.

Cause 1: Bluetooth Headphones Delayed What You Heard

This is the single most common cause of “my voice is late,” and it is not a recording bug at all.

Wireless audio adds latency, typically somewhere between 100 and 300 ms depending on the codec and the device. If you are demoing something with sound — a video, a notification, a call — through Bluetooth headphones, you hear it late. You react to what you heard, so your reaction genuinely happens late. Recorded captured both tracks correctly; you were just responding to a delayed cue.

The same latency bites in the other direction when the headset’s microphone is the input: some devices buffer the mic path as well, and the start offset Recorded measures only accounts for when the stream opened, not for latency the device adds inside every packet after that.

The fix: for anything where your voice has to land on a specific frame, use wired headphones and a wired or USB microphone. A USB mic plus any wired earbuds removes the entire class of problem. Our audio recording best practices guide covers the rest of the input-device decision.

If you must use a wireless headset, record a deliberate reference at the top of the take: click something that makes an audible sound and say a short word at the same instant. You will be able to see the gap in the editor and decide whether it is acceptable before you record twenty minutes on top of it.

Cause 2: You Switched Audio Devices Mid-Recording

Plugging in headphones, joining a call that grabs the default output, or letting a headset auto-connect halfway through a take all change the audio path while the capture is running.

On Windows, system audio is captured from the output endpoint through WASAPI loopback. If the active output device changes, the stream the capture was attached to is no longer the one producing sound. On macOS, system audio comes from ScreenCaptureKit alongside the video, which is more resilient, but the microphone stream can still be interrupted when its device disappears.

The fix: set the input and output you intend to use before you open the recording controls, then leave them alone. Quit the apps that reassign audio devices on their own — conferencing clients are the usual offenders, and they often grab the default device the moment they launch. The platform-specific setup is in how to screen record on Mac with internal audio and how to screen record on Windows 11 with audio.

Cause 3: Dropped Frames Made the Picture Fall Behind

This is the cause behind drift that grows as the recording goes on, and it is the one people least expect.

Audio capture is cheap and almost never falls behind. Video capture is expensive. When your machine cannot keep up — a saturated CPU, a nearly full disk, a sync client churning in the background — frames get dropped. Audio keeps running at real time, so the gap between what you hear and what you see widens over the length of the take.

You can usually recognize it without any tooling: the recording feels slightly choppy in the busy stretches, and the desync is worst in exactly those stretches.

The fix: reduce the capture load rather than hunting for an audio setting. Close the browser windows and sync clients you are not using, record to an internal SSD with plenty of free space, and prefer window or area capture over a full 4K monitor. Our performance optimization guide walks through finding the actual bottleneck, and the broader symptom list lives in the screen recording troubleshooting guide.

Cause 4: Only the Webcam Is Late

A camera is the slowest device in the chain. Exposure and white balance negotiation mean it can take several hundred milliseconds — occasionally more than a second — before the first usable frame arrives, long after the screen capture has started.

Recorded measures this too: when the camera track is saved, the app records how far after the screen capture the webcam actually began and stores it with the other offsets, so the webcam overlay is placed against the right part of the timeline rather than simply starting at zero.

Where it still shows up is at the very top of a take. If you start talking immediately, your first word may land before the camera has produced a frame, and no amount of alignment can recover footage that does not exist.

The fix: give it a beat. Start the recording, breathe once, then start talking. You get a clean lead-in to trim and the camera is fully running by the time your face matters. This is the same habit that makes recording in takes workable.

Cause 5: Your Voice Sounds Doubled

If your voice has a short slap or an echo behind it rather than a delay, the tracks are aligned and the problem is that your voice was captured twice.

With system audio enabled and sound playing through speakers, your microphone picks up the room — including the speakers. The same sound then exists on two tracks that cannot be perfectly coincident, because one travelled through air. Mixed together, the result is a hollow, phasey double. In a recorded call, it is worse: the far end’s voice arrives both through the system audio capture and through your mic.

The fix: wear headphones whenever system audio is on. If a take is already recorded this way, open the editor, select the affected segments, and mute or lower the track you do not need — Recorded keeps per-segment System Audio and Microphone volume controls precisely because the two streams stay separate all the way to export.

Cause 6: The Video Inside Your Recording Was Already Out of Sync

Sometimes your narration is perfect and the clip you were demoing is the thing that drifts.

A screen recorder captures whatever the display is actually showing. If the browser or player you were recording dropped frames while its own audio kept playing, the recording faithfully preserves that desync. Recording a video call has the same property: the client’s own jitter buffering is baked into the pixels.

The fix: verify in the source, not in the recording. Play the clip at full size, confirm it holds sync in its own player, and give the player a moment to settle before you start capturing. If it drifts there, it will drift in the capture.

Checking Sync in Thirty Seconds

Do not discover this after a long take. Record a short test and look at it properly:

  1. Start a recording with the exact inputs you plan to use. The live level meters next to the mic and system audio controls should both move — if one is flat, fix that before anything else.
  2. Click something with an audible sound and say a single sharp word, like “now,” at the same moment.
  3. Stop, and let the editor open the recording.
  4. Look at the timeline. Recorded draws a waveform for the system audio and the microphone under the segments, so the two spikes are visible — and they should line up with each other and with the frame where the click happens.
  5. Scrub to that frame and step through it. If the spike and the visible click are within a frame or two, you are clear.

Thirty seconds here is cheaper than re-recording a tutorial. It also tells you which fix you need: a constant gap means a device latency problem, a clean start that decays means a performance problem.

What To Do With a Take That Is Already Off

Be honest about what is recoverable.

A doubled or echoing take is recoverable: mute the redundant track per segment and the problem disappears. A late webcam at the top of the take is recoverable: trim the opening. A take where the picture dropped frames is not meaningfully recoverable — the missing frames were never captured, and stretching the audio to fit would pitch or stutter it. A take where you reacted to delayed Bluetooth audio is not recoverable either, because your performance itself is late.

For the recoverable cases, the editor’s segment controls are the tool. For the rest, the honest answer is a five-minute re-record with wired audio and a quieter machine — which is almost always faster than fighting footage.

A Pre-Record Sync Checklist

[ ] Input and output devices set before opening the recording controls
[ ] Wired headphones on, especially with system audio enabled
[ ] Conferencing apps quit so they cannot grab the default device
[ ] Mic and system audio level meters both moving
[ ] Sync test recorded and checked against the waveform
[ ] Capture scoped to a window or area rather than a whole 4K display
[ ] Enough free space on the internal drive
[ ] A one-second beat before the first word, so the camera is running

The Short Version

Sync problems come from the recording environment far more often than from the recorder. The track alignment itself is already measured and applied for you; what is left is the physics — wireless latency, devices changing underfoot, frames your machine could not keep up with, and sound captured twice.

Fix those once by standardising your setup, and run the thirty-second waveform check whenever something feels off. If you are setting up from scratch, download Recorded and make that sync test the first thing you record.

Happy recording!