Closed Captions

• Use closed captions as they can be toggled on an off by the viewer.
• Open captions are embedded into the video and cannot be “turned off”. Avoid open captions.

Captions should be 99% accurate

• Machine-generated captions are often ~80% accurate
• Use machine-generated captions as a starting point and then manually edit

• Establish identity of all speakers (whether on-screen or off-screen) by name or meaningful identifier
  • Use (Name) or (Speaker/narrator/identifier of some kind etc.) to identify different speakers.
  • By default, Kaltura uses [Speaker:] which that is acceptable as well.
• If speakers change off screen, use “-” to indicate the change in speaker
• Double check spelling of names via graphics in the video, the video description, or own research.

Captions should have proper sentence case for readability. All uppercase can be used for screaming or shouting. Use proper syntax, punctuation, and capitalization (for proper nouns and names).

• Caption lines should end at natural linguistic breaks, such as a natural pause mid-sentence of the end of the sentence to create understandable, logical segments of speech.
• Keep subjects, phrases, and key terms in a single caption, where possible.
• If a complete sentence fits within one caption, it should. Extend the timing of the caption to match.
• Start a new caption with each new sentence.
  • Two very short sentences can be displayed in a single caption.

Each caption should display at the same time the first word of the line is spoken.

• If there is an extremely quick sequence of information, fast dialogue, and/or large blocks of jargon and terminology, start the caption slightly before the speaker is heard.
• Captions should appear for at least 1 second but less than 10 seconds.
• Captions for non-speech audio and music should appear for a maximum of 5 seconds.
• Do not display more than 2 lines per caption frame.

• Sound effects, music, and other audio should be mentioned in enclosed brackets.
  • Sound effects are any meaningful audio content that is not spoken.
  • Standard practice is to describe the sound in [sound].
    • [doorbell ringing]
    • [dog barking]
    • [ambient street sounds]
  • If music is known, provide [title by artist].
    • For unknown instrumentals, describe briefly, for example: “[soft piano music]”.
  • Music with lyrics, introduce the song with [title by artist] and then transcribe lyrics in sync with the video within a symbol, for example: “#song lyrics to a song#”.

• You may ignore “um” “ah” “uh-huh” etc.
• A relevant pause can be noted with “…”.
• A notable change in thought can be marked by “–”.
• Do not censor or edit the audio content. Transcribe it as is.
• Mark inaudible or muffled audio with “[Inaudible]” or “[Unclear]”.
• Include proper punctuation and syntax.

• Transcribe “four and a half kilometers” as “4.5km”, “9 o’clock” as “9PM”, “three dollars” as “$3”, etc.
• Numbers one to ten can be written as words (one, two, etc.) except when used in units of measurement or currency.
  • Numbers above 11 should be written as digits.
  • Numbers over one million should be captioned as “1 million”, “3 billion” etc.
• Use % in place of “percent”.
• Dates:
  • A speaker saying “the second of June twenty fourteen” should be captioned as “2 June, 2014”.
  • A speaker saying “June the second twenty fourteen” should be captioned as “June 2 2014”.