Make the virtual-character video readable without the voice.

Editorial analysis

Accessibility begins in the storyboard

A character who points silently at three products while the labels flash behind captions creates a problem no transcript can fully repair. Reserve a caption-safe region in every shot. Keep essential text on screen long enough to read, name visual choices in narration or captions, and avoid placing the only clue in a subtle facial change.

For virtual characters, also check whether lip movement, eye motion and rapid overlays compete with reading. The face can remain expressive without making every corner of the frame move. Accessibility is part of the performance design, not a file attached after export.

Source record

Captions carry more than words

W3C's explanation of WCAG 2.2 Success Criterion 1.2.2 says prerecorded synchronized media should provide captions for audio content, with a defined exception for clearly labeled media alternatives to text. It explains that captions include dialogue, speaker identification and meaningful non-speech sound. It also says captions should not obscure relevant visual information.

YouTube says its automatic captions can misrepresent speech because of pronunciation, accents, dialects or background noise and tells creators to review and edit them. It also lists overlapping speakers, multiple languages, poor audio and unsupported languages among possible problems. Automatic generation is therefore a draft input, not final editorial approval.

Evidence: W3C Web Accessibility Initiative [s1] · YouTube Help [s2]

Practical application

Build the caption source before export

Use the approved script as the starting transcript, then conform it to the final edit. Add speaker labels when identity is not visually obvious and bracket only sounds that carry meaning: a door slam that interrupts the character, a phone alert that changes the plot, or music whose stop signals a reveal. Do not caption decorative sound so densely that it crowds the dialogue.

Maintain a caption glossary for character names, invented places, sponsor products and pronunciation-sensitive terms. When dialogue changes, update the script, audio and caption cue in one ticket. A caption file copied from an earlier cut is a common-looking deliverable with the wrong timings.

Practical application

Run four viewing passes

First, watch muted with captions: can a new viewer follow the story and identify speakers? Second, listen without looking: are essential visual actions described or otherwise available? Third, view at phone size: do captions cover hands, products or lower-face acting? Fourth, use the platform player: are line breaks, timing and language labels still correct after upload?

Worked example: the character rejects a red jacket with a silent head shake, while captions cover the jacket. The fix is not a caption saying “[shakes head]” over the object. The editor moves the caption, holds the shot and adds “Not the red one” to the approved dialogue, preserving both action and meaning.

Editorial analysis

Give accessibility defects owners

Log transcript errors to editorial, timing and placement to post-production, visual-description gaps to the director, and platform-file failures to publishing. Name a reviewer who can block release. “Captions present” is not the same as captions checked.

Keep the final caption file, language, reviewer, review date and platform version with the video package. Recheck after any trim, dub or music replacement. This procedure supports a more accessible delivery; it is not a conformance certification, which requires a broader review of the experience.