Audio Description for Brand Video: The Missing Track

Captions get shipped on almost every brand video now — it's routine, expected, barely a line item. Audio description is the accessibility track that didn't make the same jump, and it's not because it matters less. It's because it used to require a specialist writing a timed script against a stopwatch and a separate voice booking, for a feature most teams assumed was niche enough to skip. Neither half of that excuse holds up as well as it used to.

What audio description actually is

It's not a second caption track and it's not a transcript. WCAG 2.2's Success Criterion 1.2.5 requires audio description for all prerecorded video content in synchronized media — a spoken narration track, in the same human language as the video, describing actions, characters, scene changes and on-screen text that aren't already conveyed in the existing audio. Captions solve the "I can't hear the dialogue" problem. Audio description solves a different one: "I can't see the product being held up, the on-screen price, or the logo that appears with no one saying its name out loud." A brand video that's entirely dialogue with nothing meaningful happening visually technically clears the bar without an extra track — the criterion explicitly exempts that case — but that's a narrow exception, not the default state of a product ad or a brand story built around visual beats.

The constraint that actually makes it a craft

The requirement isn't "narrate the video" — it's narrating it inside a specific gap. Description has to fit during existing pauses in dialogue, using standard audio description technique, which means the description script is written against whatever silence the video already has, not against a fresh timeline you control. A fast-cut ad with wall-to-wall voiceover and no pauses is a genuinely harder case than a slower brand story with natural breathing room — the gap you're writing into is fixed by the edit, and the description has to be written to fit it, not the other way around.

The legal backdrop, briefly

This isn't purely a best-practice nicety at this point. The European Accessibility Act (Directive 2019/882) covers accessibility of audiovisual media services and has applied since 28 June 2025 — a binding requirement in the EU market, not a voluntary guideline. WCAG 2.2's SC 1.2.5 sits at Level AA, the tier most organizational accessibility policies and a growing number of regulations reference as the baseline. Neither of these singles out brand video specifically — they're written around "prerecorded video content" generally — but a product ad or a brand story is squarely inside that scope, not a special exemption.

Where the script actually comes from

This is the part that's changed the cost equation more than any regulation has. Writing an audio description script from scratch means someone watching the finished video repeatedly, noting every visually-conveyed beat, and timing each description against the available gaps by hand. Versely's analyze_video does the first half of that pass automatically: it extracts sample frames, describes the overall style and format, returns per-timestamp beats, and reads on-screen text — which is, functionally, the raw material an audio description script is built from, already broken into a timestamped structure rather than a video a writer has to watch cold. Adding include_transcript layers in the spoken dialogue too, which is exactly what a description writer needs to see next to the visual beats in order to find where the actual gaps are.

Writing the description itself

The beats and on-screen text from analyze_video are the raw material, not the finished script — someone still has to turn "0:04 — logo appears, bottom right" into a description that reads naturally in the gap available at 0:04, and trim it to fit rather than running long into the next line of dialogue. The craft is compression: describing only what's necessary to understand the video without the visuals, in as few words as the available silence allows, which is a genuinely different writing task than an ad script optimized to persuade.

The voice and mix step

Once the script is written and timed, this is ordinary Versely territory. generate_speech turns the description text into narration — and it's worth deliberately dialing the delivery toward neutral and unobtrusive using style_instructions, since an audio description track competing for attention with the main read defeats its own purpose; this is one read where a flat, objective, unhurried delivery is the correct choice rather than the expressive one you'd want for the ad's main voiceover. attach_audio_to_video in mix mode then layers that narration onto the video's existing audio without removing anything — the mechanism that lets the description sit inside the existing pauses rather than replacing the video's original sound.

The common, expected delivery pattern is publishing two versions rather than forcing description onto every viewer: the standard cut, and a separately exported audio-described version — the same pattern streaming platforms use with a toggleable "audio description" audio track. A brand video is rarely delivered through a platform with a built-in track-selection UI, so in practice that means two exported files rather than one file with a switchable track, published and linked as the accessible alternative rather than baked into the only available version.

A Versely walkthrough: from raw cut to a described version

1. Extract the beats. analyze_video against the finished brand video, with depth set high enough to catch on-screen text and include_transcript: true so the dialogue and the visual beats are both in front of you.
2. Write the description script. Turn each meaningful visual beat into a short description sized to fit the gap it falls in — product held up, on-screen price, logo appearance, scene change — skipping anything already covered by the existing dialogue.
3. Generate the narration. generate_speech with the timed script and style_instructions set toward neutral, objective delivery — deliberately distinct from the main ad's voice.
4. Mix it into a separate export. attach_audio_to_video in mix mode, against a copy of the finished video, layering the description into its existing pauses without touching the original audio.
5. Publish both versions. The standard cut as the default, the described version linked or labeled as the accessible alternative — Versely's voice-over tooling and the underlying add-voiceover-to-video mechanics cover this same generate-then-mix shape for a primary narration track, applied here to a second, accessibility-specific one.

Checking it before it ships

Once the described version exists, it's worth a pass through the same review step any finished deliverable gets before publishing — checking a video before it goes live catches a description that runs long into the next line of dialogue the same way it catches a caption that overflows the frame, which is exactly the kind of timing error that's easy to miss by ear alone on a first pass.

Picking a voice for the read

Not every text-to-speech model is tuned for a neutral, unhurried delivery — some are built and ranked specifically for expressive, character-forward reads, which is the wrong direction for this specific track. The best AI text-to-speech model ranking is a reasonable starting point for comparing what's available before committing a description script to a voice that fights the format it's meant to serve.

FAQ

Is audio description the same thing as captions? No. Captions convey spoken dialogue and sound for viewers who can't hear it. Audio description conveys visual information — actions, on-screen text, scene changes — for viewers who can't see it. They solve opposite problems and a video can need either, both, or neither depending on what it actually contains.

Does every brand video legally need an audio description track? It depends on the content and where it's shown, not on the fact that it's a brand video specifically. WCAG 2.2's SC 1.2.5 applies to prerecorded video with meaningful visual content in synchronized media; a video that's essentially dialogue-only with nothing separately visual to describe is exempted by the criterion itself.

Can the description just be squeezed in over the existing dialogue? No — the standard technique is fitting description into existing pauses, not layering it over dialogue that's already playing. A video with no natural gaps is a genuinely harder case, and may need an extended-description approach (pausing the video itself to make room) rather than a simple mix pass.

Do I need a different voice for the description than the main ad voice? Not a different voice necessarily, but a different delivery — audio description is meant to inform, not perform, so a flatter, more neutral read than the main ad's voiceover is the right call regardless of which voice model or identity you use for it.

Captions stopped being optional once the tooling made them cheap enough that skipping them looked like an oversight rather than a resource decision. Audio description is at the same inflection point now — the script's raw material comes out of a video you're already reviewing, and the voice-and-mix step is the same pipeline a primary voiceover already runs through.