Write captions for speech and meaningful sound

Transcribe the actual final speech and identify who is speaking when it is not otherwise clear. List non-speech sounds that affect understanding, such as a warning chime or a door closing off-screen. Do not add captions based only on an earlier script if the recorded wording changed.

W3C’s Captions (Prerecorded) guidance explains captions for prerecorded synchronized media, including speaker identity and meaningful non-speech information, with a defined exception for clearly labeled media alternatives to text. This guide’s workflow is a production aid; a captioned file alone does not establish complete accessibility.

Consider a fictional exchange about a model: Mira says “Is the channel open?” A soft latch clicks, then Tomas says “Now it is.” The draft caption rows are “Mira: Is the channel open?”, “[latch clicks]” and “Tomas: Now it is.” The click matters because it explains the state change between the two spoken lines.

If the latch action is already unmistakable visually, the caption still needs to represent meaningful audio information rather than simply duplicate every movement. A decorative whoosh on a title may not need the same prominence as a warning or off-screen event. Decide from the sound’s contribution to understanding, and have an appropriate reviewer confirm the treatment.

Align each caption with its corresponding speech or sound and keep phrase groups readable. Do not reveal Tomas’s response before Mira finishes asking the question. Separate speaker changes clearly using names or another reliable convention; color alone should not carry the distinction.

Reserve a caption zone that avoids the channel and latch labels. If the diagram occupies the bottom of the frame, recompose it or select another consistent caption position. A caption that hides the mechanism it describes creates a new comprehension problem. Check the longest caption against the smallest expected playback view.

Watch muted with captions and ask whether the question, latch event and resolution remain understandable. Then compare captions against the audible track for exact wording, speaker identity and event timing. Review language-specific punctuation and line breaks with someone qualified to assess them.

If text competes with an on-screen heading, shorten or remove redundant display copy while retaining the caption content. For a separate sound-off adaptation, essential visuals may need additional explanatory labels, but those labels should not silently replace a complete caption track. Record the final transcript version, caption timing and inspected export together so later dialogue edits trigger a fresh review.