How should brands caption video? A practical workflow

Captions, subtitles and on-screen copy do different jobs

W3C’s media-accessibility guidance defines captions as synchronized text for speech and the non-speech audio needed to understand the content. In its terminology, subtitles translate spoken audio into another language. Markets often use the words differently, so the brief should describe the actual deliverable rather than rely on a label.

A designed headline, name strap or campaign message inside the edit is not a caption. It does not replace the soundtrack, and a caption track should not simply duplicate every visual label. Treat visual copy, same-language captions and translated subtitles as separate content layers with separate approvals.

Build captions into the production plan

Leaving captions until the final export creates avoidable framing and approval problems. W3C recommends including accessibility requirements in project requirements, RFPs and contracts. For a brand team, that means assigning ownership for transcription, terminology, translation, timing, QA and platform upload before the edit is locked.

The script and storyboard should allow room for captions, name straps and other essential graphics. When one shoot must supply several aspect ratios, the vertical and horizontal framing plan should include caption-safe areas. Space for readable text is a composition decision, not a gap to find at the end.

Five decisions to put in the brief

• The source language and required translations
• Destination platforms and aspect ratios
• Selectable captions, burned-in captions or both
• Approved spelling for people, products and brand terms
• Owners for transcription, language review and final approval

Choose open or closed captions by distribution

Closed captions are a separate text track viewers can switch on or off. They can support multiple languages, remain editable and allow compatible players to apply user preferences. Open captions are burned into the picture and cannot be removed.

Brands rarely need to choose only one. A website or YouTube upload can preserve a clean master with selectable captions, while autoplay social placements, presentation screens or channels without dependable caption-track support may need a burned-in version. The most useful archive keeps all three: the clean master, the open-caption export and one caption file per language.

Write for meaning, reading rhythm and the frame

W3C’s WCAG 2.2 explanation says captions include not only dialogue but also relevant music, sound effects, laughter, speaker identity and location. The aim is not to annotate every noise. It is to preserve the information a viewer needs when the audio is unavailable.

Edit speech without rewriting the speaker

Not every hesitation needs to remain, but polishing a sentence into something the person never said can alter tone or meaning. Check names, figures, URLs and specialist terms separately. A single missing negative can reverse the message.

Let the viewer read in time

Bring cues in and out close to the spoken line. Break at natural phrase boundaries, not arbitrary character counts, and avoid forcing paragraphs into one frame. Identify speaker changes where the picture does not make them obvious. Timing should respond to language density, pace and movement, not one mechanical duration applied to every clip.

Protect important visual information

W3C notes that captions should not obscure relevant information. If a lower third, product detail, direction or offer condition occupies the bottom of the frame, move captions to another safe position. Test contrast over the actual shots, not only over a neutral editing canvas.

Use automated captions as a draft, not approval

Speech recognition removes a great deal of manual work, but it does not own the final wording. YouTube’s official guidance warns that pronunciation, accents, dialects, background noise and overlapping speakers can produce errors. It tells creators to review and correct automatic captions.

A dependable pass is simple: generate the draft, compare it with the audio, verify proper nouns, repair punctuation and speaker changes, watch the timing, then test the published result on the destination platform. Automation handles the first pass; a person protects meaning, brand language and accessibility.

A caption delivery workflow for brand teams

YouTube’s supported-file guidance recommends SRT and SBV among the basic starting formats and requires plain UTF-8 for SRT. Format support is only one part of delivery: the filename, language code, matching video version and approval status need to travel with the asset.

A practical handoff includes the clean master, channel-ready open-caption exports, an SRT for each language, the approved transcript and a short manifest stating which file belongs to which platform. That package prevents every future cut or localization from restarting the same work.

If you want to plan the brief, edit versions and caption delivery as one production system, explore our content and video services or tell us about the project.

Common questions

Are automatic captions good enough to publish?

No. They can be a useful first pass, but a person should check names, brand and product terms, punctuation, speaker changes, timing and errors that alter meaning.

Should captions be burned into the video or supplied as a separate file?

Keep a clean master and a separate caption file whenever the channel supports selectable captions. Also create an open-caption version for placements where a caption track is unavailable or muted viewing is central to the experience.

Do captions need to include music and sound effects?

Include non-speech audio that is needed to understand the content, such as a relevant music cue, door sound, applause or speaker change. Decorative sound does not need to be described indiscriminately.

Is an SRT file a complete caption delivery?

SRT is a widely supported basic format, but the file still needs the correct language label, UTF-8 encoding, complete timecodes and a test on the destination platform.