Why exhibit media needs text versions
Sound in a gallery is unreliable. Some visitors are deaf or hard of hearing. Others cannot hear over a crowd, are standing near a speaker that has been turned down, or are learning in a second language. Text versions of speech and sound make the content usable for all of them. They also make it searchable, easier to translate and easier to update.
Three related items are worth separating:
- Captions are timed text displayed with the video, containing speech and also meaningful non-speech sound, such as music or a door slamming.
- Subtitles are timed text of the speech, usually translated, and generally assume the viewer can hear other sounds.
- Transcripts are full text of the speech and often a description of important sounds and visuals, provided as a separate document or page, not timed to the picture.
Each has its use, and a project often needs more than one. Sign language options and hearing loops serve other needs, and they sit alongside text versions and do not replace them.
Accuracy comes first
Automatic speech recognition can produce a first draft quickly, but drafts typically contain mistakes, especially with names, place names, specialist vocabulary, accents, overlapping voices and poor recordings. Treat any automatic draft as a rough start and have a person check every line against the audio. Errors in a caption are not just cosmetic; they can change meaning, and visitors who rely on the text have no other source.
A sound approach is:
- Write or obtain a script if one exists, then compare it to what was actually said, since speakers rarely follow scripts exactly.
- Transcribe what is spoken, including important false starts only when they matter.
- Check names and technical terms against a reference list.
- Review the full text against the final edit, not an earlier cut.
For oral history and other recorded speech, the transcript also serves the archive, as discussed in oral histories.
What to include
Speech should generally be captured faithfully. Light cleaning is common, such as dropping a repeated “um”, but avoid rewriting what people said or smoothing out dialect so heavily that the speaker’s voice is lost. Where in doubt, keep the words.
Also include meaningful non-speech sound, in brackets or another consistent style:
- Speaker identification when the speaker is not visible or changes, such as a name before their line.
- Sound effects that matter to the story, such as [applause] or [machinery running].
- Music, described when it affects mood or meaning, such as [slow violin music].
- Languages other than the main one, indicated where the speaker switches.
- Silence or pauses only when they carry meaning.
Skip sound that adds nothing, like room hum that is not part of the story. Decide on a convention, such as brackets and lowercase, and apply it throughout.
Timing, length and layout
Captions are read in short bursts while the viewer is also watching the picture, so timing matters.
- Keep lines short. Two lines at a time is a common limit, with line breaks at natural phrase boundaries rather than mid-phrase.
- Allow enough reading time. Show each caption long enough to be read, and not so long it lingers over the next scene. Reading speed guidance exists in published accessibility guidelines, and slower reading speeds are appropriate for general audiences, children and second-language readers.
- Match the speech. The caption should appear close to when the words are spoken. Large offsets confuse viewers.
- Break at scene changes when possible.
- Position carefully. Captions should stay clear of on-screen text, faces and key action. A video with a graphic lower third may need captions moved.
- Make them readable. A plain, high-contrast typeface, enough size for gallery viewing distance and a solid or semi-solid background help; the principles of color contrast apply directly to caption styling.
Offering an option to turn captions off is common on web and app video. In galleries, many teams show captions by default, because a visitor who needs them will often not know how to turn them on.
Open and closed captions
Open captions are part of the picture and always visible. Closed captions are a separate track that a player can show or hide. Open captions are the dependable choice for simple gallery players, since they require no special support from hardware. Closed captions are flexible and suit web delivery and multiple languages. Whichever you choose, test on the real player, as support varies.
Keep caption files in a standard text-based format that can be edited and reused, and keep them with the video project files so they are updated when the film is re-cut.
Transcripts
A transcript should be easy to find. On a website, place it right next to the media. In a gallery, it might be a printed sheet, a screen option, or accessible on a phone via a code, using the approach described for visitor phones. A good transcript includes:
- A title and a note of what the media is.
- Speaker names.
- Descriptions of important sounds.
- Descriptions of key visuals that are not clear from speech, which helps deaf-blind readers and anyone who cannot watch the video. This links to the practice in audio description.
Use clear headings and short paragraphs. A long transcript can be broken up at natural topic shifts.
Screen reader users can read text transcripts with their tools, a point developed in the article on screen readers, so use real text and not pictures of text.
Translation and multiple languages
Captions can be translated into other languages, creating subtitles. Plan for this when timing the originals, since translated text is often longer and needs more screen time or smaller chunks; see exhibit translation. Translators should be given the timing file and the picture, not just a list of lines, and a native reader should check the final result against the video.
Check before opening
Watch the finished film with the sound off and then again with captions on. Ask someone who relies on captions to review them if possible. Check that the text matches the final edit, that no caption overlaps important visuals and that the files play correctly on the exhibit hardware. Schedule updates whenever the video changes, since outdated captions are a common and avoidable fault.