Transcript

A transcript is the programme as a document. Speaker labels, what was said, and in the descriptive form the things that were only shown, all laid out to be read at the reader’s own pace instead of the player’s. Nothing about it is synchronised, which is exactly what makes it useful: it can be skimmed, searched, quoted, translated, printed, and read on a braille display, none of which a timed track allows.

The comparison with the other two alternatives is worth getting straight. Captions are time locked to the picture and serve a viewer who is watching without hearing. Audio description is time locked the other way and serves a viewer who is hearing without seeing. A transcript is locked to nothing, and it is the only one of the three that works for a reader who is both deaf and blind, since it can be rendered as braille. That is why WCAG treats a full text alternative as the substitute route at level A for prerecorded media, and why a captioned video with no transcript still leaves a real audience out.

The plain version carries dialogue only. The descriptive version merges in what the description track would have said, so a reader who never plays the video still learns that the letter was unsigned. Timestamps are the small feature that makes a long one usable: printed beside each passage, they let a reader jump the player to that moment, and a player that scrolls the transcript as it plays turns the document into a navigation control. Speaker labels do the same work labels do anywhere, which is to keep a wall of text from becoming a puzzle.

The commercial arguments are real and slightly beside the point. A transcript is indexable where audio is not, it gives a search engine and a language model something to read, it is what a reader in an open-plan office actually uses, and it is the cheapest source for clips and quotes. Produce it once from an accurate captioning pass rather than twice: a reviewed caption file already contains the words, and merging cues into paragraphs with speaker names is mostly editing. Unreviewed automatic speech recognition is not a transcript, it is a first draft with the names spelled wrong.