What is audio description

Audio description is narration for the picture

“Narration added to the soundtrack to describe important visual details that cannot be understood from the main soundtrack alone.”

WCAG 2.2, definition of audio description

Captions put the sound into words for people who cannot hear it. Audio description does the opposite job: it puts the picture into words, spoken, for people who cannot see it. WCAG says it covers actions, characters, scene changes, on-screen text and other visual content. You will also see it called video description, described video or descriptive narration; they are the same thing.

W3C's guide to description says it serves people who are blind and others who cannot see the video well enough, and gives a one-line example: “Pat opens a small box, looks at a diamond engagement ring, and cries.” Without that sentence, a listener hears a box open and someone crying, and has to guess why. Description only applies to video; a podcast has nothing to describe.

For a small business the stakes are less dramatic and more concrete: the price that flashes up on screen, the button a demo clicks, the phone number on the end card. If it is only shown, a customer who cannot see it does not get it.

Standard and extended description

Standard description fits into the pauses that are already there: a narrator speaks while nobody else is talking, and the video runs at its normal length. Extended description pauses the video to make room, then lets it carry on. WCAG says the extended kind is only used when the sense of the video would be lost without more description and the pauses are too short.

Extended description has its own criterion, 1.2.7, at Level AAA, so it is not required to meet Level AA. That does not make it a luxury. A 30-second ad with a voice-over from first frame to last has no pauses at all; standard description has nowhere to go, and the honest choices are to extend it or to recut. Where each kind sits on the timeline is drawn as a timing diagram on our audio description services page.

When WCAG asks for it

Read together, the two give you a choice at Level A and take it away at Level AA. W3C's Understanding note for 1.2.3 spells it out: at Level A you may provide either audio description or a full text alternative; to conform at Level AA, 1.2.5 requires audio description, so a transcript alone no longer meets it. If you described the video for 1.2.3, you have already met 1.2.5 too.

Two cases sit outside that pair. Description of live video is not required to meet WCAG, though W3C notes that people who are blind still need it; a presenter who says what is on each slide covers most of it. And a recorded video with no sound at all falls under 1.2.1 instead, which accepts either a text description or an audio track.

Level AA is the level that matters in practice. The DOJ Title II web rule for state and local governments names WCAG 2.1 Level AA, which includes 1.2.5. Governments with a population of 50,000 or more face 26 April 2027, and smaller ones and all special district governments 26 April 2028, after DOJ's April 2026 interim final rule moved each date back one year; check ada.gov for the current dates, and see our Title II web rule guide. For private businesses DOJ has no detailed web regulation and points to WCAG as guidance. The full set of video rules, captions included, is in video caption requirements. This is general information, not legal advice. For your own situation, ask a lawyer.

When a video needs no description at all

WCAG's definition carries its own exemption: where the soundtrack already carries all of the video's information, no description is needed. W3C's example is a video that is only a person talking. A founder speaking to camera in front of a plain wall usually passes. Add a name caption, a price on screen, a chart or a silent demo, and it usually does not.

The test is not whether the video has pictures; every video does. It is whether a listener misses anything they need. Lighting, the colour of a shirt and the room behind the speaker rarely matter. A phone number on the end card, a step shown and not said, or a product that is held up and never named always do. W3C suggests telling users when a video needs no description, so a listener knows nothing is missing.

Integrated description: write it into the script

“Description of the visual information can be seamlessly integrated by the speakers as the video is planned and created, and you don't need separate description, thus there is no additional cost.”

W3C Web Accessibility Initiative, Description of Visual Information

This is the cheapest way out of 1.2.5 for most small businesses, and W3C puts it first in its own order of choices. Its guide to planning audio and video says that for presentations and instructional videos, the best way to handle description is not to need it at all. Its kitchen example makes the point in one line: “Whip the mixture until it looks like this” leaves a listener stuck; “Whip the mixture until the oil, vinegar, and spices are well combined” does not.

W3C's tips list what to watch for: title text at the start, links and email addresses at the end, speakers' names shown in text, and the text on any slide. It does not have to be read word for word, as long as all of its information is in the audio. Every “here”, “this one” and “like so” in a script is a place where the picture is doing the talking.

A worked example: three lines, before and after

A bike shop is filming a 40-second clip on how to book a repair online. The first draft of the script reads naturally to anyone watching, and says almost nothing to anyone listening. Here are three of its lines, and the same lines with the description written in.

The after version is a few words longer per line and needs no narrator, no description track and no special player. It also helps people you were not thinking of: someone listening with the screen locked, or watching on a phone too small to read the end card. Check the finished cut once with your eyes closed; if you can still book the repair, the description is done.

The text alternative: a descriptive transcript

The other half of 1.2.3 is a document WCAG calls an alternative for time-based media: the speech, the sounds and the visual information, in the order they happen, as text. It reads like a screenplay, and WCAG adds a warning about exactly that: the shooting script only counts if it is corrected to match the final edit.

W3C's transcripts guide calls this a descriptive transcript and explains who it is for: people who are both deaf and blind, who read it on a braille display, and anyone who takes in text better than video. It also notes that once you have captions and description for Level AA, a descriptive transcript is easy and inexpensive to make from them. Publish it as HTML on the same page as the video.

Text alternatives for still images work on the same idea at a smaller scale; our alt text examples show how to say what an image is for in one line.

How to get description for a video you already have

For a video that is already finished, W3C's description guide names two routes. Write the descriptions and put them in a timed WebVTT file that the player reads aloud, or write them, record a narrator, and mix them into a new audio track or a second, described copy of the video. Many organizations, it adds, choose to outsource the job.

The first route uses the same WebVTT format as a caption file, so if you write the descriptions yourself, a caption editor that exports WebVTT can hold the timed lines. Our closed captioning software comparison covers those editors.

Which route works depends on your player more than on the vendor. W3C's media player guide lists Able Player and video.js as players that can read description from a text file, and Able Player, OzPlayer and video.js (with a plug-in) as players that can play a separate description audio track. If your player can play only one soundtrack, that usually means ordering a second, described version of the video.

Whoever writes it, the brief is the same, and W3C sets it out: describe what matters to understanding, plainly and without comment, in the present tense; voice it in a style distinct from the people in the video; and place each description with or just before what it describes, never after. Send the describer your product names and the script, and read their script before it is recorded.