Match an accurate timed text track to the final video edit, configure its language and purpose, then verify real cue playback and delivery rather than treating an uploaded file as completion.
A caption file is useful only when its words, timing and delivery match the video a visitor actually plays. Work from the final media edit, review the text, connect the intended track and turn captions on in the public player. A successful upload is just one step.
This workflow uses a fictional paper-folding demonstration. The sample words and times illustrate a small track; no actual video, player or caption delivery was tested for this guide. Use your real recording for every timing and accuracy decision.
Separate timed captions from nearby page text
A paragraph underneath a video can describe its purpose, but it does not follow the dialogue during playback. Timed captions appear in coordination with the media. A transcript gives a separate reading route; it should not be mistaken for an enabled track.
W3C’s caption guidance covers speech and relevant speaker or non-speech information. Automatic transcription needs checking. For the folding example, review a negative instruction carefully: losing “do not” can reverse the intended action.
Decide who reviews subject-specific terms, names and numbers. Preserve meaningful sounds where they matter to understanding, rather than describing every incidental background noise. Check the recording itself; a transcript supplied from an earlier script can differ from what was actually said.
Lock the media version before synchronising
Record the final video filename or stable identifier, duration and revision alongside the track. If an introduction is shortened after timing the captions, later cues may all drift. A small edit is still an edit when the track’s timestamps depend on it.
For the fictional demonstration, suppose the first title card is removed. Do not fix only the first cue and assume the rest remain aligned. Identify the change’s position and review every affected section against the revised playback.
Keep the editable original transcript and timing project recoverable. A replacement video needs an explicit decision: reuse the existing track after checking it, retime it or create a new version. Avoid an untraceable collection of files all called final-captions.vtt.
Prepare a small valid WebVTT file
MDN’s WebVTT format reference specifies UTF-8 text, a text/vtt media type, the WEBVTT header and separated timed cues. An end time must follow its start; timestamp milliseconds use a dot. Renaming a subtitle file with different syntax to .vtt is not a conversion.
This original miniature example is for the fictional clip. Its timings are illustrative and must not be reused for another recording:
WEBVTT
00:00:01.200 --> 00:00:04.400
Fold the top corner toward the centre.
00:00:05.000 --> 00:00:07.500
[Paper rustles]Save as plain text rather than a word-processing document. Inspect the saved extension and encoding. Keep a simple dialogue track easy to audit, with clear cue boundaries; overlapping cues are not universally forbidden, but accidental overlap can obscure this straightforward instruction.
Preview short sections at normal playback speed. Break text at meaningful points and allow enough reading time for the actual language and content. Do not impose an unexplained universal character count as if it guaranteed readability for every video and viewer.
Connect language, purpose and a useful label
MDN’s track reference distinguishes kind, source, language and the user-facing label. Captions can convey relevant non-speech audio; subtitles serve a different language-understanding purpose. A metadata track is not a visible caption substitute. Use the correct language tag, required for subtitles, and a label the viewer can recognise.
For the paper-folding clip, English captions and a translated subtitle track should have distinct understandable choices. Do not label two different languages “Captions” and expect the visitor to discover which is which by starting playback.
Check the player’s supported default choice and the effect of a viewer’s own preferences. The existence of a track in saved markup does not mean it is currently displayed. Test selecting it deliberately, and confirm selecting another track changes the text shown.
Use the actual player’s track controls
In the native WordPress Video block, the documented Text tracks controls allow a .vtt track and its label, source language and kind to be managed. The ordinary Add/Remove caption control is the text beneath the video, a different feature. Inspect the installed version and save the intended track settings.
An external video service may manage captions in its own player account instead. Find the maintained workflow for the player actually embedded. Adding a local text file near an unrelated embed does not attach it to that provider’s playback.
Reopen the saved public page logged out. Confirm that it contains the intended video revision and available track choice. Keep any written explanation available if the media cannot play; do not make the entire instructional answer depend on the player loading.
Diagnose a missing track at the delivery layer
Inspect the requested track URL and actual response, not only its Media Library presence. A login page, missing file or stale response can prevent useful captions. Check the returned content and content type through the site’s supported diagnostic process.
The track source normally needs the document’s origin unless the media parent uses crossorigin. That attribute alone does not grant permission: the maintainer must also review the relevant server CORS response and delivery configuration. Use the CORS diagnosis guide for that separate responsibility, without disabling browser protection globally.
If a CDN or file replacement serves an older track, compare the actual published response with the intended version. Do not repeatedly retime correct text to compensate for an old file still being delivered.
Review playback beyond the first cue
Separate a text-correction pass from a timing pass. A reviewer can first compare a section’s wording with the recording, then watch it again without stopping to assess its reading rhythm. Record the cue and proposed correction precisely so the editor can change the right version without disturbing already approved sections.
- Play from the beginning and enable the intended caption track.
- Check speech, technical terms, negatives and meaningful sounds against the audio.
- Seek into the middle and near the end; verify the expected cues appear.
- Review transitions, pauses and any edited section for drift.
- Change language choices and confirm their names and text agree.
- Repeat in supported desktop and phone players, including fullscreen where used.
- Check caption readability without covering essential visual instruction.
Retain the media/track version, tested combinations and corrections. The video-page guide covers the broader loading and playback decision. This track workflow is accepted when the correct text appears at the correct moments and remains a usable choice for the viewer, not merely when a file exists.
Sources and verification
Editorial review 2026-10-15. Source check dates are recorded separately. Notes describe what each source supports and its limits.
View sources
- MDN: WebVTT text-track formatChecked UTF-8 text/vtt files, WEBVTT header, cue separation and start/end timestamp structure.
Limits: A syntactically plausible sample does not establish accurate transcription, timing or full media accessibility. - MDN: track elementChecked Track purpose, source, language, label and the same-origin/crossorigin distinction.
Limits: Correct markup does not establish successful cross-origin delivery or a particular player’s presentation. - W3C WAI: Captions/SubtitlesChecked Meaningful speech, speaker and non-speech information and review of automatic captions.
Limits: Caption quality must be reviewed against the actual audio; captions alone are not a complete accessibility assessment. - WordPress: Video blockChecked Timed Text tracks and editing label, language and kind, distinct from the ordinary caption below a video.
Limits: Installed-version and player output need checking; an external provider may use a separate track-management workflow.