Video Captions: Keep the Editable Track with the Final Cut

The finished video has elegant captions, and the editor sends over an MP4. A week later, the creator notices a misspelt product name. They correct the separate caption file, but the words visible in the video remain unchanged. The handover contained two different forms of text, and only one was updated.

For an independent creator, that distinction belongs in the production brief. A video export, a timed caption file and an editable project serve different purposes. Keeping them associated with the same final cut makes corrections and future editions easier to trace.

Name the caption format you are commissioning

The W3C Web Accessibility Initiative’s caption guidance distinguishes closed captions, which viewers can show or hide, from open captions, which remain visible. It describes captions as synchronised text conveying speech and the non-speech audio information needed to understand the content.

Text burned into the picture is one way of delivering always-visible captions. A separate timed track gives a supporting player a different resource to display. Ask the editor which version is being delivered and where the editable caption text will be kept. “Captions included” leaves those questions unresolved.

My proposed handover would preserve the reviewed timed text even when a particular destination needs a burned-in export. That is a production recommendation, rather than a claim that every platform accepts the same caption format or offers the same controls.

Keep the text and timing in an editable file

YouTube’s supported caption-file documentation explains that these files contain words and time codes specifying when each line appears. For SubRip files, with the extension .srt, YouTube supports basic versions in plain UTF-8 and does not recognise style markup.

This means a caption file that uploads successfully is not evidence that a particular font, colour or placement has travelled with it. For a YouTube delivery, inspect the result in YouTube’s player. For another destination, check that destination’s documented requirements instead of assuming the same behaviour.

In a fictional handover, an editor might supply a video without burned-in text, its reviewed English caption file and a separate export with captions rendered into the picture. The editable project would be retained where agreed. These are proposed deliverables, not files produced or tested for this article.

The useful distinction is between changing the text resource and changing the rendered picture. Correcting an SRT does not, by itself, replace letters already rendered into an existing MP4. For that burned-in version, return to the appropriate editable source, make the correction and produce a revised export.

Attach the captions to a particular cut

Imagine a fictional ceramic artist preparing a seventy-two-second demonstration. The first six seconds contain an introductory title with no spoken material. The editor creates and reviews captions, then removes that introduction for a shorter edition.

A spoken cue originally beginning at twenty seconds now begins at fourteen seconds, assuming there are no other timing changes. That is simple timeline arithmetic, not a test of any editing application. The caption file must be checked against the revised cut; the earlier approval cannot establish its new timing.

A real edit may be less uniform. Removing a passage in the middle affects material after that cut differently from material before it. Ask the editor to review the actual revised sequence, rather than applying an unexplained adjustment to every cue.

MaryChuks.com’s guide to checking video timing across the whole clip explains why an opening check can miss later mismatches. For caption handover, the same production discipline is useful: review the text against the exact exported picture and sound.

Review an automatic draft before approving its derivatives

YouTube’s automatic-caption guidance says machine-generated quality can vary and instructs creators to review and correct transcriptions. It identifies factors including accents, dialects, mispronunciations and background noise. Automatic availability is therefore a starting point for review, rather than evidence that the wording is approved.

For the ceramic demonstration, the artist would check the product name and technical vocabulary against the actual recording. The editor would also review the timed text with the finished demonstration. Only the reviewed version should become the basis for the separate track and any burned-in export.

This avoids an unnecessary split in the process: one person correcting the track while another renders an older draft into the picture. Record which text version was used for each export. If the audio wording changes, revisit both the transcription and its timing.

Make the handover legible without opening everything

Use a simple edition reference in the production notes. For example, the fictional delivery might identify demo-cut-03.mp4, demo-cut-03.en.srt and demo-cut-03-burned.mp4. A filename helps associate the items, but the relationship still needs verification.

• Video without burned-in captions: identify the approved picture-and-sound version.
• Timed caption file: state the language, cut reference and review status.
• Burned-in export: identify the caption revision rendered into its picture.
• Editable source: record where the agreed project and caption-editing materials are retained.

This list is a proposed handover convention. It does not guarantee that a project will open elsewhere or that a caption file is supported by a destination. Those capabilities need their own checks in the agreed software and delivery route.

Check the delivered playback, including both caption states

For a version using selectable captions, open the delivered video in its intended player and check the relevant track. Turn it on and off. Confirm that the selected language matches the reviewed file and that the text remains aligned at representative moments throughout.

If the video also contains burned-in text, inspect what happens when the player’s caption track is enabled. Two independently displayed versions could duplicate or obstruct one another. Decide the intended delivery arrangement and test it explicitly.

For the burned-in export, inspect the final image at its intended viewing size, including any crop used for a shorter social edition. Check that the words remain readable and do not cover the demonstration. Record the tested player, version and file; these proposed checks are not a claim that the fictional video has passed accessibility testing.

The final handover should leave the next editor able to answer three questions: which cut is this, which caption text belongs to it, and which delivered versions need changing if a correction arrives? Preserving those relationships makes captions a maintainable part of the publication, alongside the picture and sound.

Sources

• W3C WAI: Captions/Subtitles — updated 17 September 2024.
• YouTube Help: Supported subtitle and closed caption files — undated guidance, checked 30 October 2026.
• YouTube Help: Use automatic captioning — undated guidance, checked 30 October 2026.