How to Fix Subtitles That Are Out of Sync

When subtitles are out of sync, the first temptation is to drag every line until the opening sentence looks right. That can make the problem worse. Subtitle timing errors usually fall into two different patterns: every cue is shifted by about the same amount, or the error grows as the video continues. Those patterns require different fixes.

Diagnose the pattern before editing

Play the video and inspect three points: the first spoken line, a line near the middle, and a line near the end. Note whether each caption appears early or late, and by roughly how much.

If the first, middle, and last cues are all late by about two seconds, you have a constant offset. If the first cue is correct but the final cue is many seconds late, you have drift. If only one section is wrong, look for an incorrect cue or an edit in the video rather than shifting the entire track.

Write down the evidence before making a change. That simple step gives you a way to undo a bad adjustment and prevents a local problem from becoming a global one.

Fix a constant offset

A constant offset is the straightforward case. Select all subtitle cues and move them earlier if they appear late, or later if they appear early. Keep the duration between each cue’s start and end time the same. You are moving the track, not rewriting the captions.

After the shift, test the first line and a line several minutes later. If both align, check that the first cue does not begin before the video’s useful audio and that the final cue still ends before the video ends. A negative start time or a cue that extends past the video can create a second problem during export.

Be careful with files that contain multiple subtitle languages. Shift only the language track you intend to repair, or apply the same measured correction to every track if they were created from the same timing source.

Fix gradual drift

Drift means the subtitle and video do not share the same timeline. Common causes include using subtitles from a different cut, changing video speed, or mixing frame-rate assumptions during export. A global shift cannot solve this because the needed correction changes from the beginning to the end.

First confirm that the subtitle file belongs to the exact video version. Compare the duration, opening, inserted ads, removed pauses, and ending. If the video was trimmed after the subtitle file was created, repair the cues around the edit points or regenerate the track from the new audio.

If the start is aligned and the end is consistently early or late, a time-scale adjustment may help. Use the first and last reliable cue as anchors, apply the correction across the range, and then spot-check the middle. Do not stretch a file blindly when the video has several cuts or speed changes.

Check frame rate and export settings

Frame-rate errors are more common with older broadcast-style files and converted footage, but they can still appear in modern workflows when one tool interprets timestamps differently. A subtitle file stores time offsets, while some editing environments also reason in frames. If the video was converted from one frame rate to another, inspect the export settings and the tool’s subtitle import options.

WebVTT cue timestamps use hours, minutes, seconds, and milliseconds. The file format is not a guarantee that every editor will interpret a source timeline identically. Export a short test section after changing frame or time settings, then compare the beginning and end of that section before processing the full video.

Repair a single bad section

If only one caption is wrong, do not shift the whole track. Split a long cue, correct the start or end, and inspect the cues immediately before and after it. Overlapping cues, a missing gap, or a cue that lasts far too long can make viewers perceive the next line as wrong even when its timestamp is correct.

Check spelling and segmentation at the same time. A line that displays the right words at the wrong beat is a timing issue; a line that contains words from the next sentence is usually a transcription or segmentation issue. Keeping those diagnoses separate makes review faster.

Verify the fix from beginning to end

Watch the corrected video at normal speed. Test with sound on to check alignment and with sound off to check that captions remain long enough to read. Review a busy scene, a speaker change, a number, and the final call to action. If the file will be uploaded to a platform, upload a private or unlisted version and check the rendered captions there too.

YouTube’s caption workflow lets you upload a timed file or edit captions in Studio. The uploaded file is not the final proof; platform rendering can expose line, timing, or language-selection issues that did not appear in the editor.

When regeneration is faster

Repair is worthwhile when the transcript is accurate and the timing problem is small. Regeneration is usually faster when the video has been heavily recut, the audio has changed, or many cues contain errors. Keep the old file as a backup, create a new track from the final audio, and compare the result rather than layering uncertain fixes on top of one another.

The key is to diagnose first. Constant offset, progressive drift, local errors, and export mismatches look similar in a paused frame but behave differently across the timeline.