By ToolBrivio Editorial · Documentation-based guide. No hands-on results are claimed. We earn no affiliate commission from links on this page. Research approach · Link disclosure.
The words are correct, but the subtitles appear before the speaker says them. You fix one caption, then find another wrong several minutes later. This often happens when the subtitle file belongs to an earlier edit of the recording.
Before changing dozens of timecodes, establish which file is the reference. The finished video or audio should determine subtitle timing. A transcript, an Audacity project and an exported recording can each represent a different stage of the work.
Basis: This guide uses official Audacity, Happy Scribe and Descript documentation checked on October 6, 2026. The examples are hypothetical. We did not run a synchronization test, and the diagnosis patterns below suggest what to investigate rather than prove a particular cause.
First check whether the media itself is synchronized
Play the final video without subtitles. Do visible mouth movements, claps or other clear events match the sound? If the audio is already out of step with the picture, correct that media problem before retiming captions. Otherwise, you could align subtitles to an audio track that will move again.
Also confirm that the player or editor has loaded the intended subtitle file. A similarly named older export can make a correct revision appear broken. Keep the final media and its matching subtitle export together, with a version identifier in both filenames.
Compare three anchors instead of one caption
Choose a short, clearly spoken phrase near the beginning, one in the middle and one near the end. Avoid overlapping voices or a phrase that repeats frequently. For each anchor, note when the subtitle starts and when the matching words are heard.
The pattern matters more than the first discrepancy. Use the following as a decision aid, then confirm it against the edit history.
| Observed pattern | Possible explanation to investigate | Correction approach |
|---|---|---|
| Similar offset throughout | A shifted start or added/removed lead-in. | Consider the same time adjustment for affected cues, then check later anchors. |
| Error grows gradually | A duration or playback-speed difference between versions. | Check the media versions and speed settings; a single offset will usually be insufficient. |
| Timing matches, then suddenly changes | A cut, insertion or rearrangement partway through the edit. | Inspect that boundary and realign the affected section. |
These explanations are possibilities. A mixed edit can produce more than one pattern, so do not turn the table into an automatic rule.
Pattern 1: every caption is late by roughly the same amount
Imagine subtitle starts at 120, 600 and 1,200 seconds, while the matching speech starts at 118, 598 and 1,198 seconds. At those three anchors, the captions are consistently two seconds late.
A two-second earlier adjustment is a reasonable candidate. For example, a cue running from 00:02:00,000 to 00:02:03,000 would become 00:01:58,000 to 00:02:01,000. Moving both endpoints preserves its duration. Changing only the start can leave an unnecessarily long cue or introduce overlap.
Use a tool that supports the required adjustment; this guide does not establish that every editor offers a global shift. Inspect early cues separately so none acquire a negative timestamp. Check all three anchors again and look for captions that now collide with neighboring cues.
Practice a fixed offset with two subtitle files
Download the before file (SRT), corrected file (SRT), and three-anchor worksheet (TXT). These original practice cues have no accompanying recording. They demonstrate a two-second earlier shift, not verified synchronization with real speech.
Cue 1 changes from 00:00:12,000 --> 00:00:14,500 to 00:00:10,000 --> 00:00:12,500. Each endpoint moves earlier by two seconds, so its 2.5-second duration is unchanged. Cues 2 and 3 follow the same rule.
Compare all three cues in a text editor: the wording, numbering and durations should remain identical. For a real project, fill the worksheet with heard speech starts near the beginning, middle and end. If those differences are not approximately the same, this fixed-shift exercise is not an adequate repair. Recheck against the final media before delivery.
Pattern 2: the mismatch gets larger as the recording continues
Suppose the same subtitle anchors are 120, 600 and 1,200 seconds, but the speech occurs at 119, 595 and 1,190 seconds. The difference grows from one second to five, then ten. Subtracting two seconds from everything would improve some captions and leave others wrong.
Check whether someone changed the speed or stretched the recording after subtitles were created. Audacity's trim-and-stretch documentation explains that speed changes alter a clip's duration. That makes a speed difference worth investigating; it does not establish that speed is the cause in this example.
Compare the source and final media durations and review any intentional speed change. If the finished media is correct, realignment against that version is often safer than repeatedly applying small offsets to an older subtitle file. After realignment, check sentence boundaries and cue endings as well as starts.
Pattern 3: captions work until one particular edit
If timing is good before a cut and wrong immediately afterward, examine the cut itself. Removing three seconds while pulling all later audio earlier would make unchanged later captions three seconds late. Inserting three seconds would produce the opposite direction.
Audacity distinguishes edits that leave a gap from edits that move later material. Its basic editing guide documents those choices and notes that Delete behavior can be configured. Check which operation was actually used rather than assuming that deleting audio preserved timing.
Do not shift the already correct opening section. Repair captions crossing the cut boundary individually: a sentence may have been shortened, removed or joined to another passage. Then realign the later section, checking for additional edits that create another timing change.
Silence removal can change more than one boundary
Automatic silence shortening deserves special attention. Audacity's Truncate Silence documentation explains that the effect shortens duration. Its independent-track option can also pull selected tracks out of alignment when each is processed separately; applying the same selected-track edits avoids that particular divergence.
If subtitles were created before silence shortening, there may be many separate timing changes. Treat the finished media as the reference and realign against it. For multitrack work, first confirm that the speakers still line up correctly with one another and with any video. Synchronizing selected audio tracks inside Audacity does not automatically retime an external video or subtitle file.
Make the repair in a timeline, then export the matching version
Happy Scribe's timeline guide documents moving and resizing subtitle clips and adjusting timecodes against the waveform. Its subtitle tutorial also describes Sync realignment. Review the current workflow and your plan's export options before choosing a delivery format.
If many cues no longer match, our recommendation is to create or realign subtitles against the final media, then review the text again. A clean new timing pass still needs proofreading; it can introduce recognition errors or different line breaks. Keep the corrected older text as a reference rather than assuming regeneration preserves every correction.
When bringing captions into a video editor, use the same media edit and matching timing reference. Happy Scribe's DaVinci Resolve import guide addresses caption import and timecode matching. An import can be technically successful while the captions still belong to another cut.
Descript's subtitle documentation distinguishes SRT/VTT text-track exports from captions burned into a video. A separate subtitle file can be replaced; a burned-in correction requires a newly exported video. Check which deliverable you are repairing.
Check the exported result in its destination
Reopen the final export or uploaded preview and revisit your beginning, middle and ending anchors. Inspect every repaired cut boundary, captions containing removed words, overlaps and the final cue. Confirm that you are checking the newly exported file.
For future revisions, finish cuts and speed changes before final subtitle alignment. If the media changes afterward, record what changed and where, then repeat the timing check. Keeping a matched media-and-subtitle pair is a small habit that prevents a correct caption file from becoming the next version's synchronization problem.
Related help
For the file structure, read Your first SRT file, explained. For the delivery process, use our subtitle handoff checklist. If you are choosing an audio editor, read the Audacity 4 review.