Audio description

Some of what a video says is only ever shown. A character reads a letter without saying a word about it, a diagram fills the frame while the presenter says “as you can see here”, a number appears on screen and is never spoken. Audio description is the second narration track that says those things out loud, written to fit the silences the dialogue leaves and voiced between the lines rather than over them. The constraint is the whole craft: a describer has a few seconds of room at a time, has to spend them on what actually carries the story, and has to get out of the way before the next line lands.

When the gaps are not long enough, the standard track has to drop information, which is what extended audio description exists for. There the playback pauses so the description can finish, then resumes, and the running time of the video grows. WCAG asks for description of prerecorded video at level A, where a full text alternative is allowed as a substitute, and again at level AA, where description is required outright. Extended description sits at AAA, and pausing playback is only workable where the viewer controls the player rather than sitting in a cinema.

The three text and audio alternatives are often confused, and they answer different questions. Captions turn the soundtrack into synchronised text, for a viewer who cannot hear it. Audio description turns the picture into sound, for a viewer who cannot see it. A transcript is the whole programme as text, dialogue and description alike, readable without playing anything, which is the only route that works through a braille display. A described video with no captions and a captioned video with no description each leave out a different audience.

Writing them is a specialist job, closer to editing than to copywriting. Describers work in the present tense, name what is on screen without interpreting it, keep character names consistent with the dialogue, and resist filling every silence, since a pause can be the point. Delivery is a recorded human voice or synthesized speech from a timed text file, and the web mechanism is a second audio track the player mixes in. Cheaper than all of it: shoot and script so the audio carries the meaning already, say the number rather than only showing it, and the description track has less to rescue.