Exported Video With a Fixed Audio-Video Offset? Timeline, Offset Measurement and Export-Layer Diagnosis Guide

2026-11-20 | GPU Repair | 26 minutes | HK GPU Fix Editorial Team

Exported Video With a Fixed Audio-Video Offset? Timeline, Offset Measurement and Export-Layer Diagnosis Guide

Key takeaway: When an exported video is out of sync by the same amount from start to finish, with normal pitch and speed, the error is a fixed offset rather than drift. Measure the error in milliseconds at both ends, compare the timeline preview with the export, and re-export with software encoding. That separates the timeline, the export layer, and the player.

An exported video can be out of sync in a very specific way: the sound is early, or late, by the same amount from the first frame to the last. A clap at the start is off by the same number of milliseconds as a spoken word at the end. The pitch of the voices is normal, and nobody sounds sped up or slowed down. Only the position of the sound has shifted, as a whole, against the picture. That combination — a fixed error, the same at both ends — is the whole problem this guide addresses, and the combination itself is the first piece of evidence.

The instinct in this situation is usually to call the problem drift, or to drag the audio track sideways on the timeline until the finished file looks right. Both instincts are wrong for this shape. Drift grows: it starts correct and gets worse towards the end. A fixed offset does not grow. And moving the audio track on a timeline that already plays correctly does not repair the export — it damages a correct timeline to match an incorrect finished file. The first job is therefore to measure, not to move anything. Measure the error at the start, in the middle, and at the end, and only then decide which layer is at fault.

This guide covers only a narrow shape. Five conditions must all be present. First, the finished export — from Premiere, DaVinci Resolve, or Final Cut, as an MP4 or MOV — has audio and video out of sync, with the sound consistently early or consistently late, in one direction throughout. Second, the error is fixed: the head, the middle, and the tail are off by the same number of milliseconds. Third, pitch and speed are normal: the sound has not changed pitch and has not become faster or slower; it has only shifted position. Fourth, the error is baked into the finished file: every player playing the same finished file at the same timecode hears the same error. Fifth, the state of the timeline must be stated plainly: whether the timeline itself plays correctly and only the finished file is wrong, or whether the timeline is wrong too. The two are different layers and must not be mixed together. If the error grows towards the end of the file, the shape is drift, not a fixed offset. If the pitch changes as well, the shape is a speed error, not a fixed offset. If only one player shows the error, the shape is a player problem until proven otherwise.

A fixed offset is not drift, and the two must never be swapped

A fixed offset and progressive drift are measured differently, and they are repaired differently, so the two words must never be used interchangeably. A fixed offset is the same number from start to finish. Drift starts correct and grows worse as the file goes on. A speed error changes pitch and speed together with the timing. The outside sources consulted for this guide agree on this division, and they agree on how to test it.

The wefixsound guide states that a fixed offset, with audio consistently early or late through the whole video, is the simplest kind of sync problem, and that shifting the audio track back by the measured offset is what addresses it. The toolsforfilm guide states that measuring at two points settles the question: if the two points are off by the same number of frames, the error is a fixed offset, caused by a timecode break or an import error, while an offset that grows with the length of the clip means the speed is wrong. The hollyland guide states that audio early or late by a fixed amount, unchanged through the whole clip, is an offset and not drift. The chatcut guide states the same test in the other direction: the same gap at both ends is an offset, and a gap that grows larger is drift. The sonilo guide states that a fixed offset means the lips and the sound are off by about the same amount, and that the check should be made at the start, the middle, and the end.

The consequence of confusing the two is a wrong repair. A fixed offset is addressed by shifting position (after measuring the offset at the export/compensation layer — not by moving audio on a timeline that already plays correctly). Drift is addressed by changing speed. A speed error is addressed at the sample-rate layer. Measuring only one point in the file cannot tell them apart. The head, middle and tail measurement in Step 2 is therefore the step that locks the shape in place, and it must not be skipped.

A gap handed over twice, and received by nobody

The fixed offset of an exported finished file is the one shape in the library that two guides hand over and nobody receives. That double handover is the reason this guide exists, and it should be stated plainly rather than left as an implication.

One guide hands the shape over in its comparison table, writing that an equal error at both ends with normal pitch is a fixed offset, not its own subject, and directing the reader to the guide about a television. Another guide writes in its boundary section that a whole-file fixed offset, of the kind seen when sound goes to a television, is a different kind of problem, and pushes it out of its own scope. But the television guide narrows its own opening section to a graphics card connected to a television by HDMI, and what it describes is television picture-processing delay and the sound path to the television. It takes only the HDMI-to-television kind. It does not take the exported finished-file kind. A further guide states in its opening section that a fixed offset between lips and sound is not its subject, and excludes it outright.

Put together, the position is this: two guides say a fixed offset is not theirs, the television guide takes only the television kind, and another guide explicitly refuses it. The exported finished file with a fixed offset is therefore claimed by no guide at all. That is a gap. This guide fills that gap and nothing else. It does not re-cover the television case, the sample-rate case, or the drift case. Those remain with their own guides, and this guide routes the reader to them whenever their shape appears.

How the export layer can produce a fixed offset

When the timeline plays correctly and only the finished render is wrong, the fault sits in the export layer, and there are several ways the export layer can produce a fixed offset. The hollyland guide describes this as a less common but particularly frustrating situation: the timeline looks correct, and the problem appears only after a full render. It lists hardware encoding errors, an export sample rate that does not match the sequence, and Match Source presets that do not necessarily carry the audio settings across correctly. Its stated comparison is to turn off Hardware Encoding in the export settings and export again with software encoding.

The wefixsound guide adds the codec and container layer: when the editing software is in sync and the export is not, the codec settings should be checked — the audio and video export frame rates and the sample rate must match — along with variable frame rate material, the container format, and B-frames. The toolsforfilm guide names a fourth root cause, import offset error: some editing software adds a fixed offset when importing some media types, above all MXF audio from professional recorders and MP4 from consumer cameras, so the whole clip ends up off by the same fixed amount without drifting. A timecode break works in a similar way: the error is fixed within a stretch of the file, and it changes at the break point, so measuring at several points is what reveals it.

The export settings layer also includes the sample rate. The hollyland guide states that 48 kHz is the video standard, and that an export sample rate that does not match the sequence is one of the causes to check. The pitch and speed of the sound must be confirmed as normal before this layer is blamed, because a sample-rate error that changes speed belongs to a different shape. The three candidate causes named by the sources — hardware encoding, import offset and timecode break, and codec and container settings — sit at different layers. They do not contradict each other, and this guide does not rank them. The diagnosis must first divide the timeline layer from the export layer, and only then ask which export cause is in play.

A correct timeline and a wrong export do not mean the timeline is wrong

The sonilo guide states the most important warning in this whole area: when only the export is wrong, check the export frame rate, sample rate, channel mapping, player, and packaging, make a short test, and do not damage a correct timeline to suit one playback path. That warning matters most here, because it is the error most easily made with this shape.

If the timeline plays correctly and the finished file does not, the export layer is wrong, not the timeline. Moving the audio track on the timeline to make the finished file look right is damaging a correct thing to match a wrong thing. The timeline must be left alone until the timeline-versus-export comparison in Step 3 has been made. Only if the timeline itself is wrong does the timeline become the place to work. The same warning applies to Sync Offset. Sync Offset is a comparison and compensation measurement: it records how many milliseconds the sound is off by. It is not a repair, and writing it as a repair turns a measurement into a cause. The offset must be written down first. What to do about it depends on which layer the measurement points to.

The five suspect layers

LayerWhat it meansDecisive comparison
Timeline and sequenceThe timeline itself is already off: a clip moved by accident, a link broken, an import offset, or a Sync Offset set wronglyTimeline preview comparison: a correct timeline points to the export layer, a wrong timeline points to the sequence layer
Export encodingA hardware encoding error, a codec or container packaging fault, a Match Source preset, or B-framesSoftware encoding comparison, plus a re-export reproducibility comparison, plus a container change comparison
Sample rate and frame rate settingsThe export sample rate does not match the sequence, or the material frame rate does not match the sequence, with pitch and speed normalExport settings comparison at 48 kHz and at the sequence frame rate, plus a pitch-normal confirmation
Player and decodingOnly some players show the error, so the finished file itself may be correctEvery player plus a second device comparison
Source timecodeAn import offset from MXF or consumer MP4 material, or a timecode break after which a stretch of the file is off by a fixed number of framesSegmented measurement comparison: a different offset before and after the break point means a timecode break

Each layer has its own comparison, and no single comparison covers all five. The head, middle and tail measurement separates a fixed offset from drift and from a speed error. The timeline preview comparison separates the sequence layer from the export layer. The player comparison separates the finished file from the playback path. The order in the seven steps below is the order that wastes the least time.

Seven diagnostic steps

Step 1 — Define the shape

Confirm the five shape conditions: the finished file is out of sync, the error is fixed, pitch and speed are normal, the error is baked into the finished file, and the state of the timeline has been stated. Write down the baseline before changing anything: the editing software and version, the source of the material — which device recorded it, which container, which codec — the sequence settings for frame rate and sample rate, the export settings for hardware or software encoding, container, and sample rate, and the direction of the offset — whether sound is ahead of picture or picture is ahead of sound — together with the approximate number of milliseconds. A diagnosis built on a vague description such as "it is out of sync" cannot be checked later, because it does not say how far off, in which direction, or whether the timeline was correct.

Step 2 — Measure the error at the head, in the middle, and at the tail

This is the most important single step in the whole guide, and it must not be skipped. Find a moment in the finished file that can be recognised by eye and by ear — a clap, a burst of sound, a lip shape — and measure the error near the start. Measure it again near the middle, and again near the end, in milliseconds. If all three points are off by the same number, the shape is locked in as a fixed offset, and the diagnosis continues to Step 3. If the error grows steadily from the head to the tail, stop this guide: the shape is progressive drift, and a different diagnostic path applies. If the error jumps suddenly between segments — the head and the middle agree, but the tail is off by a different fixed number — the shape is a timecode break, not drift: the error is fixed within each stretch and changes at the break point, so note where the break falls and continue to Step 3. If the pitch changes as well, stop this guide: the shape is a sample-rate error, and a different diagnostic path applies. The measurement must use milliseconds as the unit. Words such as "a little off" or "very off" cannot be compared between the points, and they cannot be written down as a result.

Step 3 — Compare the timeline preview

Play the same stretch on the timeline inside the editing software, and check whether the timeline plays correctly or is also off. A timeline that plays correctly, with a finished file that does not, points to the export layer. A timeline that is also off points to the sequence or source layer, and the export settings should not be investigated further. This step separates the timeline layer from the export layer. The comparison takes the same form as the guide about exported files with some silent clips, but what is measured here is different: that guide measures whether there is sound or not, while this guide measures how many milliseconds the sound is off by.

Step 4 — Compare players

Play the same finished file in VLC, in the editing software, and on a second device — another computer or a phone — and write down the direction and the number of milliseconds for each one. The same offset in every player points to the finished file itself. An offset in only some players points to the player and decoding layer, and the finished file must not be blamed directly. Wireless playback delay can make in-sync media look late, so the comparison must use wired headphones or built-in speakers, not Bluetooth headphones. Bluetooth delay is a property of the playback path, not of the file, and it cannot be used to measure an offset in the file.

Step 5 — Measure the direction

Once the error is confirmed as a fixed offset, write down its direction: whether the sound is ahead of the picture, or the picture is ahead of the sound. Use the lips as the reference. Either the mouth moves and the sound arrives afterwards, or the sound arrives and the mouth moves afterwards. The direction is part of the measurement, and it must be recorded together with the number of milliseconds. The method of measuring direction by lips is the same kind of measurement used for a television, but the receiving end is different: that case is a television, and this guide does not use its television-processing comparisons.

Step 6 — Test whether a re-export repeats the fault

Export again with the same settings, and check whether the offset is the same number, in the same direction, every time. The same number every time is a systematic problem. A different number every time points to the encoder layer. Add two further comparisons. First, export again with software encoding, turning off Hardware Encoding: if the offset disappears, the fault points to the hardware encoding layer. Second, change the container — MP4 to MOV — and export again. This step is a comparison test, not a recommendation to change settings permanently. Re-exporting only proves which layer is at fault. It does not mean the original file must always be exported with software encoding from now on. A single re-export that happens to come out correct proves only that the fault did not occur that time.

Step 7 — Stop if the fault is still undivided

If the fault still cannot be divided after these steps, or if the finished file also crashes players, MediaInfo cannot read the file, or the offset is different every time and grows larger with each export, stop and bring the written record to Room 1209, Dali Plaza, Shek Mun for inspection. The record should list the editing software and version, the material source and codec, the sequence frame rate and sample rate, the export settings for hardware or software encoding, container, and sample rate, the measured offset in milliseconds at the head, in the middle, and at the tail, the direction of the offset, the timeline preview result, the player comparison results, and the re-export results. The telephone number is 2186 8010 and WhatsApp is 6558 6806. Do not delete or overwrite the original file; it is the comparison material. The fee for inspection is given in the closing section below.

Boundary cases

A finished file whose pitch has changed, with speech faster or slower, is a sample-rate problem: speed and pitch move together, and it is not a fixed offset. A finished file that is correct at the start and grows more out of sync towards the end is progressive drift, at the variable-frame-rate layer. Lips out of sync while watching a television connected by HDMI is a television problem: the receiving end is television picture-processing delay, not an exported finished file. A recording file in which two sources disagree with each other — one source off by one amount and another by none — is a source-clock problem, with two sources drifting against each other, not a whole file with one fixed offset. An exported finished file in which some clips are silent and others are sounding is a different symptom dimension: that guide measures whether there is sound or not, while this guide measures how many milliseconds the sound is off by. Audio over HDMI or DisplayPort that cuts out intermittently is a link re-handshake problem, not a fixed offset. A finished export with sound and no picture is the reverse direction, at the video-track layer. A recording file that is silent for the first few seconds and then gains sound is a different time shape, at the start-silence layer: this guide is about a fixed offset through the whole file, not about silence at the start.

Six common misconceptions

  1. "Out of sync means drift." Wrong. A fixed offset and progressive drift are measured differently: equal errors at both ends mean a fixed offset, while an error that grows towards the tail means drift. Without measuring the head and the tail, the word "out of sync" says nothing about which one it is.
  2. "Move the audio track on the timeline and it is fixed." Wrong. If the timeline plays correctly and the finished file does not, the export layer is wrong. Moving the audio track on the timeline damages a correct timeline to match a wrong finished file. The timeline layer and the export layer must be divided first, and the timeline must not be moved directly.
  3. "Sync Offset is a repair." Wrong. Sync Offset is a comparison and compensation measurement. Writing down how many milliseconds the sound is off by is a diagnosis. Treating it as a repair turns a measurement into a cause.
  4. "Exporting it again fixed it." Wrong. Re-exporting is a comparison test. The hardware encoding, sample rate, or container layer must be divided before there is a direction for repair. A single re-export that comes out correct proves only that the fault did not occur that time.
  5. "One player shows it out of sync, so the finished file is broken." Wrong. Every player plus a second device must be compared. An offset in only some players is a player or decoding problem, not a fault in the finished file. Bluetooth headphones carry their own delay and must not be used to measure an offset.
  6. "A fixed offset must have been made during editing." Wrong. An import offset from MXF or consumer MP4 material, a timecode break, or the export encoding layer can all produce a fixed offset, and the timeline itself may be completely correct. The editing must not be blamed until the timeline preview comparison has been made.

Sources and what they do and do not establish

The chatcut guide, "How to Sync Audio and Video (and Fix Drift)", states that the same gap at both ends is an offset, that a gap growing larger is drift, that a sudden jump is a further kind, that a player comparison should use another player with built-in speakers or wired headphones, and that an OBS recording can carry a fixed microphone-to-camera buffering offset from the start.

The wefixsound guide, "How to Fix Audio Sync Problems in Video", states that audio sync problems run from fixed offset to progressive drift, that a fixed offset with audio consistently early or late through the whole video is the simplest kind, that sync trouble after export with the editing software in sync is a codec-related problem, and that DaVinci Resolve audio sync offset adjustment and sliding the audio track in Premiere are manual offset methods. Its codec and container checks — audio and video export frame rate, sample rate, variable frame rate material, container format, and B-frames — are the basis for the export layer in this guide.

The toolsforfilm guide, "What to Do When Your Audio and Picture Are Out of Sync in the Edit", names four root causes — a sample-rate mismatch needing a speed correction, a timecode break needing a position correction, a frame-rate conflict needing reinterpretation, and an import offset error in which some editing software adds a fixed offset when importing some media types, above all MXF audio and consumer-camera MP4 — and states the two-point test: the same number of frames at two points means a fixed offset, while an offset growing with clip length means the speed is wrong.

The hollyland guide, "How to Fix Premiere Pro Audio Not Synced with Video", states that audio early or late by a fixed amount, unchanged through the whole clip, is an offset and not drift, that a sequence or frame-rate mismatch shows from the first frame with the same error throughout, and that a loss of sync only after export has causes including a hardware encoding error, an export sample rate that does not match the sequence, with 48 kHz as the video standard, and Match Source presets that do not necessarily carry the audio settings across correctly. Its stated comparison is to turn off Hardware Encoding and export again with software encoding. The guide also notes that some GPU drivers with H.264 hardware encoding can introduce drift in files longer than 20 minutes; that note is about drift, not about a fixed offset, and it is not used here as evidence for a fixed offset.

The sonilo guide, "How to Sync Audio and Video: Complete Guide", states that a fixed offset means the lips and the sound are off by about the same amount, that a reliable event should be aligned and then the start, middle, and end checked, that when only the export is wrong the export frame rate, sample rate, channel mapping, player, and packaging should be checked with a short test, and that a correct timeline must not be damaged to suit one playback path. It also states that Bluetooth playback delay is not a file sync problem.

The vgkami guide on fixing OBS dropped frames and audio desync for charity streams states that audio trouble should first be divided into fixed, progressive drift, and intermittent jump, that a recognisable lip-to-sound moment near the start and near the end of a test recording should be compared, with equal gaps at both ends meaning a fixed error and a growing gap meaning drift, and that Sync Offset only compensates a stable delay and is not a general repair. The clipa.studio guide on fixing screen-recording audio out of sync on a Mac lists three patterns — wrong from the first second, correct at the start and drifting later, and correct in preview but wrong in the output file — and states that a fixed offset usually points to a capture-settings problem and that shifting a whole audio track only repairs one point when the offset grows.

The sources disagree about which cause of a fixed offset is most prominent: hollyland emphasises hardware encoding, toolsforfilm emphasises import offset and timecode break, and wefixsound emphasises codec and container settings. Within its own section, hollyland states that a hardware encoding error is the most common cause — that is that site's wording, and this guide does not adopt it as its own ranking. The three are about different layers, not about each other, and this guide does not adopt any of them as its own ranking. No source cited here gives a universal number of milliseconds for a fixed offset, and this guide gives none: every offset must be measured file by file. No percentage, case count, success rate, or fee beyond the site's published three-tier fee appears in this guide, because none is supported by the sources used. The fee itself is given in the closing section below.

FAQ

How do I know whether my file has a fixed offset or drift?

Measure the error in milliseconds near the start, near the middle, and near the end, using a moment that can be recognised by eye and by ear, such as a clap or a lip shape. The same number at all three points means a fixed offset. An error that grows steadily towards the tail means progressive drift. A sudden jump between segments — the head and the middle agreeing, but the tail off by a different fixed number — means a timecode break, not drift. A change in pitch as well means a speed error. One measurement cannot separate them, so all three points must be measured.

The timeline plays correctly but the export is out of sync — is the timeline wrong?

No. A timeline that plays correctly with a finished file that does not points to the export layer, not to the timeline. Moving the audio track on the timeline to make the finished file look right would damage a correct timeline to match a wrong finished file. Leave the timeline alone, and test the export layer with a software-encoding re-export and a container change.

Does Sync Offset repair the problem?

No. Sync Offset is a comparison and compensation measurement. It records how many milliseconds the sound is off by, and that measurement is part of the diagnosis. Treating it as a repair turns a measurement into a cause. What to do about the offset depends on which layer the other comparisons point to — the timeline, the export encoding, the settings, or the player.

My export is out of sync in one player but fine in another — is the file broken?

Not necessarily. Play the same finished file in VLC, in the editing software, and on a second device, and write down the offset for each one. The same offset in every player points to the finished file itself. An offset in only some players points to the player and decoding layer, and the finished file must not be blamed on the evidence of one player alone.

My re-export came out correct — does that mean the problem is solved?

No. A single re-export that comes out correct proves only that the fault did not occur that time. Re-exporting is a comparison test, not a repair: the hardware encoding, sample rate, or container layer must be divided before there is a direction for repair. Export again with the same settings and check whether the offset is the same number, in the same direction, every time — a different number every time points to the encoder layer.

Why must I test with wired headphones instead of Bluetooth?

Bluetooth playback carries its own delay, and that delay can make in-sync media look late. A measurement made through Bluetooth headphones measures the playback path as well as the file, so it cannot be trusted as a measurement of the file. Use wired headphones or built-in speakers for the player comparison, and write down the direction and the number of milliseconds for each player and each device.

If the fault still cannot be divided

Bring the written record described in Step 7 to Room 1209, Dali Plaza, Shek Mun for inspection. The telephone number is 2186 8010 and WhatsApp is 6558 6806. There is no Mong Kok shop, and there is no on-site service. The site's published three-tier fee is HK$400–780, HK$480–1,380, and HK$680–2,800, an overall range of HK$400–2,800, with the actual fee quoted after inspection. Do not delete or overwrite the original file; it is the comparison material.

Related articles

Related services

If you need Chip-level GPU repair, see these service pages: