Audio description
A narration track that explains important action, settings, expressions, and on-screen text that blind viewers cannot see.
See it
What it is
Audio description is narration of important visual information in a video: actions, expressions, scene changes, settings, and on-screen text that the original soundtrack does not convey. The description is usually placed in natural gaps between dialogue and important sound. It lets blind and low-vision viewers follow information that captions, which represent audio as text, do not cover. Boston's WGBH built the first broadcast version, the Descriptive Video Service, and launched it on PBS in 1990; the streaming version is the 'Audio Description' track sitting in the audio menu next to the language options.
Use it when visuals carry meaning that cannot already be understood from the soundtrack. WCAG 1.2.5 requires audio description for prerecorded video at level AA. A production may offer a selectable described audio track or a separate described version. When the soundtrack has no room for necessary narration, an extended-description version can pause the video to make space.
Gotcha: do not narrate every object or repeat dialogue. Describe what the viewer needs to understand the story, task, identity, or change, using specific language that fits the available gap. A synthetic voice and an AI-written draft can help production, but the result still needs a human editor to check timing, pronunciation, objectivity, and whether it talks over meaningful audio.
Ask AI for it
Create an audio-described version of this prerecorded video. First use the supplied media and transcript to mark every important visual action, speaker identity, scene change, and piece of on-screen text that the soundtrack does not already communicate. Write concise, objective narration that fits verified gaps without covering dialogue or meaningful sound, then record it with confirmed name pronunciations. Mix the narration with the original audio using FFmpeg's amix filter and export a separate described media file. If necessary information cannot fit, produce an extended-description version that pauses the picture. Add a clearly labelled 'Audio described' choice to the player, retain the WebVTT captions, and provide a time-coded review sheet for a blind human reviewer.