Free Subtitle Editor & Sync Fixer

Edit an existing subtitle track without changing the video

Use this subtitle editor to correct words, adjust cue boundaries, repair a consistent delay, or convert a reviewed basic track between SRT and WebVTT. You start with existing timed text, either pasted into the editor or opened from a local file. The result is an editable subtitle file or a plain transcript. The video itself is not re-encoded, and the editor does not produce a new video with captions permanently burned into its frames.

A cue is one text segment with a start and an end. Reviewing that small unit helps you connect a sentence to the moment it belongs on screen. Keep the wording, line breaks, speaker names, and bracketed sound notes that explain the recording. Correct an uncertain name against a reliable reference rather than replacing it with a guess. Better timing cannot make an inaccurate instruction or an incorrectly identified speaker trustworthy.

Processing happens in your browser without an account or subtitle upload to this service. Optional local video gives you a reference for checking the track, not a transcription input. There is no speech recognition, automatic audio alignment, translation, URL downloading, or automatic caption generation. If you have only a recording and no timed text, prepare a track in an appropriate transcription workflow first, then bring the resulting supported file here for editing and review.

Start from the exact video edit that will accompany the final captions. A shortened introduction, removed pause, or different take can make a perfectly valid subtitle file appear wrong. Preserve the original track and give the working version a recognizable name before making changes. This is a session-based editor with no saved browser history of your project, so download useful checkpoints and the finished file before leaving the page or closing the browser.

Open a supported source and verify what was imported

Open or paste basic SRT or WebVTT text. A subtitle file must be UTF-8, no larger than 1 MB, and contain no more than 1,000 cues. Optional local video has a separate 200 MB limit and must be playable in your browser. These are this editor’s input boundaries, not claims about the limits of every video destination. Choose a matching local review copy if the final media is too large, keeping its timeline identical.

Check the accepted cue count, first sentence, last sentence, and several names immediately after import. Malformed cues must be reported instead of quietly skipped. A failed source import should leave the previously accepted project intact, so verify which track remains on screen before continuing. Likewise, confirm that optional media has actually loaded and plays. Selecting a new filename is not enough to establish that you are reviewing the intended recording.

This editor focuses on basic text tracks. Advanced VTT STYLE blocks, REGION blocks, and live timestamp maps are refused with an explanation rather than imported with information silently removed. Supported VTT identifiers, settings, and markup need a VTT working copy when they matter. The preview remains plain text; it is not a styled-caption renderer. Keep the original advanced source and obtain a suitable basic export if your work requires features outside this focused scope.

Conversions can also lose information from an otherwise accepted track. Read and acknowledge warnings before exporting VTT metadata into SRT or TXT, or converting SRT with unsupported formatting or position metadata. The exported words may remain useful without being equivalent to the source. If accents or punctuation look damaged, resolve the file’s encoding before editing. A UTF-8 export of incorrectly decoded text cannot recover the intended characters by itself.

Revise cue text and structure while preserving meaning

Select a cue and review its start, end, and text together. Move a boundary because the spoken event requires it, not simply to make neighboring rows look evenly spaced. Preserve intentional line breaks when they help a phrase read naturally. When editing technical material, keep a short terminology reference for names, units, and product labels. A consistent approved spelling across the track is more useful than several plausible variations that a viewer must mentally reconcile.

Add a cue when existing timed text genuinely omits a spoken phrase or meaningful sound. Delete one only after checking that its content is absent from the final edit or intentionally covered elsewhere. For a split, choose a time inside the original interval and the corresponding text or line boundary. Review both resulting pieces so the operation does not duplicate a phrase, lose a speaker label, or leave one piece too brief to read.

Merge with the next cue when the combined statement belongs together and the timing still fits. Listen across the intervening gap before accepting the merge. A long silence or a speaker change may be a reason to keep the cues separate. Review any supported metadata as well: combining text does not automatically resolve conflicting identifiers or presentation settings. Structural editing is an editorial decision, not a shortcut for reducing the number of warnings shown for the track.

Navigation and export must keep a pending valid edit. If an entered boundary or another required value is invalid, correct it before proceeding; the editor should block instead of exporting an older accepted version behind the scenes. Use undo for recent successful changes, then inspect the restored cue before continuing. Saving a downloaded checkpoint after a substantial wording pass gives you a stable reference when the next pass focuses on timing rather than language.

Choose a shift or a two-anchor correction for the timing pattern

First check matching speech near the beginning, middle, and end. A similar error everywhere suggests a constant shift. An error that grows steadily can fit a linear drift correction. A sudden jump after a scene change suggests an edit mismatch that may need local repairs. The editor does not discover this pattern for you. Optional playback helps you identify it, while the transformation controls apply the explicit relationship you enter.

A shift follows new time = old time + shift for every cue boundary. Positive means later and negative means earlier. Durations and gaps remain unchanged. Preview the proposed times before applying the move, including the earliest cue. A negative result is refused rather than clipped to zero. This avoids silently shortening the opening cue to conceal a correction that does not fit the whole track.

Two-anchor retiming uses two subtitle reference times and the corresponding target-video times. Its scale is a = (v2 - v1) / (s2 - s1), its offset is b = v1 - a * s1, and each boundary becomes a * old time + b. Use distinct chronological references and matching target events. Because scaling changes durations, inspect reading quality after the change as well as checking the two anchor moments.

For an invented simple mapping, subtitle times 10.000 and 110.000 belong at video times 11.000 and 113.000. The scale is 102 / 100 = 1.02 and the offset is 0.800. Subtitle time 60.000 therefore maps to 62.000. Check that middle event independently. A linear mapping cannot repair several internal cuts or changing speed relationships. When the middle fails despite matching anchors, investigate the source versions instead of assuming the calculation has aligned the audio automatically.

Work through a three-cue repair tutorial example

This invented teaching example uses a repair tutorial with three basic cues. Cue one runs from 4.000 to 6.500 seconds and reads Turn the power off first. Cue two runs from 7.000 to 9.500 and reads Remove the cover carefully. Cue three runs from 10.000 to 13.000 and reads [switch clicks] Test the new setting. These timings illustrate the editor’s operations; they are not instructions for carrying out a real repair.

Suppose the matching video includes an extra 0.750 seconds before the demonstrated sequence, with no other timeline changes. Each cue needs to appear later by that same amount. Preview a global shift of +0.750. Cue one becomes 4.750 to 7.250, cue two becomes 7.750 to 10.250, and cue three becomes 10.750 to 13.750. Check all six boundaries rather than looking only at the first start time.

The first two durations remain 2.500 seconds each, and the third remains 3.000 seconds. The two original 0.500-second gaps also remain 0.500 seconds. That consistency is what you expect from an offset rather than a stretch. If the first instruction now matches but the last one still misses progressively, undo or reassess before adding more changes. The evidence would no longer support the assumption that an added opening was the only difference between versions.

After applying the verified shift, listen to all three cues against the video. Keep the switch sound note with the event it explains, confirm the wording, and inspect the boundaries during normal playback. Export SRT if that is the requested delivery and reopen it to verify 00:00:04,750 to 00:00:07,250 for cue one. A VTT export uses 00:00:04.750 to 00:00:07.250 for the same interval. Conversion changes timestamp notation, not the accepted timing.

Interpret review warnings without treating them as universal limits

Review warnings help locate cues worth inspecting: overlaps, out-of-order entries, long lines, short durations, and fast text. They do not all describe the same kind of problem. Two simultaneous speakers can justify an overlap, while an accidental collision can confuse the exchange. Investigate before changing it. The editor does not silently sort the source, delete overlapping content, or decide that every warning must disappear before a useful track can be delivered.

Reading speed is approximated using Unicode code points including spaces divided by cue duration in seconds. It is not words per minute or measured visual width. In the teaching example, Turn the power off first. contains 25 code points including spaces and punctuation. Over 2.500 seconds that is 10 code points per second. A constant shift leaves that value unchanged because the duration does not change. A scale correction can change it by stretching or compressing the interval.

Thresholds are editable review preferences, not fixed platform limits or accessibility certification. Consider the vocabulary, sentence structure, speaker changes, and the amount happening visually. A familiar short phrase and a dense technical explanation can demand different effort even at the same measured rate. Improve a flagged cue by checking the audio, choosing a meaningful division, or revising approved wording. Do not remove a necessary name or sound note simply to make the numerical warning disappear.

Hard invalid timing is different: a negative start or an end that does not follow its start must be corrected before timed export. After the numerical checks, watch a continuous passage at normal speed with sound, then consider what remains clear without it. Finally review the destination player, where layout and wrapping may differ from this plain preview. A clean local warning list is helpful, but it cannot replace a human check of meaning, synchronization, and final presentation.

Export timed text, a transcript, or a review handoff

Download SRT or VTT when the next step needs text synchronized with playback. Choose the format the receiving workflow actually accepts and retain a working copy that preserves supported source information. Conversion is not a promise of identical presentation. Review loss warnings for identifiers, settings, or unsupported SRT metadata before acknowledging a simplified export. Keep meaningful speaker names and sound notes visible in the resulting delivery, and inspect the output rather than assuming the preview proves complete preservation.

A plain TXT transcript is useful for wording review or a separate reading document. It is not a timed track, and exporting one does not transcribe missing speech or describe unspoken visual information. Read it independently of the video and prepare any needed verified context in the transcript workflow. Copying text can support a focused review, but keep the downloaded timed master so later changes can be reconciled with actual cue boundaries.

If a PDF report is available, use it to summarize timing changes, review checks, and next steps for a colleague. It is a review and handoff document, not the subtitle file that a player needs and not a replacement for the original media. A recipient checking sync still needs the matching video and timed export. Clearly label the report so it is not accidentally sent where a caption attachment or editable source was requested.

Reopen the timed export and compare the cue count, earliest and latest boundaries, a speaker change, and a distinctive phrase. For a lossy conversion, include a brief note explaining what was intentionally omitted. Name the files with the video revision and language so another editor can pair them correctly. Download the accepted work before leaving the session. Neither an undo history on the current page nor a review PDF should be treated as a permanent backup of the editable track.

Carry the reviewed file into the right finishing workflow

The platform contexts help you prepare material for different viewing situations, from a full tutorial to a short vertical demonstration or an interview excerpt. They are not nine guaranteed subtitle upload slots. Check the current account and publishing surface before choosing the final delivery format. Some workflows use a separate subtitle attachment; others need a video editor or native caption finishing step. The prepared timed source remains useful without implying that every destination accepts the same file.

Use the relevant platform guide for that preparation context, and the practical timing or format guide when the problem is shared across destinations. Keep one approved timed master for each distinct video edit and language. A short version with removed sentences needs a separate timing review even if it reuses wording from the longer recording. Do not overwrite the full-length master just because one delivery calls for fewer cues or a different text arrangement.

Hand the receiving person the requested file, the exact media revision, and a concise record of unresolved decisions. Mention deliberate overlaps, approved spellings, and any acknowledged metadata loss. Ask for a final playback check after the track is attached or rendered in the destination. This closes the gap between preparing valid text and confirming a useful viewing result. If the video changes afterward, return to the timing reference before treating the previous export as current.

This free editor stays separate from SMTasker account data and does not automatically import SRT into the desktop client or publish it to a platform. SMTasker uses Android phones with a local Windows or macOS client for its supported workflows. Once the subtitle handoff is complete, the five-day trial is a separate next step for exploring those account tasks. Keep the caption file and publishing responsibilities explicit so neither a conversion nor a trial signup is mistaken for an automatic native upload.