Video, audio, captions and transcripts

An automatically generated caption track is a reasonable starting point. On its own, it is frequently not accurate enough. Automatic speech-to-text commonly mishears names, technical terms, accents and background noise. WCAG’s requirement is for captions that are accurate. Merely present is not enough.

What’s required for prerecorded video, live streams and audio-only content, and why captions and transcripts serve different purposes

Content type	What’s needed	Who specifically benefits
Prerecorded video with sound	Accurate captions	Deaf or hard-of-hearing users, and anyone watching with sound off
Audio-only content (a podcast, an audio recording)	A text transcript	Deaf or hard-of-hearing users, and screen reader users who can’t play audio content usefully
Live video	Live captions where practical	Deaf or hard-of-hearing users following in real time
Video conveying information visually with no audio	An audio description or an equivalent text alternative	Blind or low-vision users who can’t perceive the visual information

Captions and transcripts are not interchangeable. A transcript alone doesn’t help someone who wants to watch the video with captions while it plays. Captions alone don’t serve someone who wants to skim or search a text version rather than watch. Video content generally benefits from having both available.

Why auto-generated captions usually need review, not blind trust

Automatic speech-to-text technology has improved significantly. It still reliably struggles with proper nouns, industry-specific terminology, strong accents, overlapping speech and background noise. That is exactly the kind of content common in a small business’s own promotional or explainer videos. Publishing auto-generated captions without reviewing them against the actual audio risks captions that are present but wrong. That can be worse than no captions at all, if a viewer relies on them and is misled by an error.

A practical, low-cost review process

Auto-generate a first-pass caption file using your video platform’s built-in tool. Then have a person watch the video with the captions on. Correct errors as they appear — names, technical terms, and anywhere the auto-caption clearly misheard something. This is meaningfully faster than transcribing from scratch. It also produces a materially more reliable result than publishing the automatic version unreviewed.

What to do for audio-only content: producing a text transcript

A podcast or audio recording published on a website needs a text transcript to be accessible to deaf and hard-of-hearing users. It separately serves screen reader users, users in a quiet environment where playing audio isn’t practical, and anyone who prefers to scan text rather than listen. A transcript can often be produced from the same speech-to-text process used for captions. Review it the same way, for accuracy.

Live streamed content, closed captions and subtitles

Live captioning is technically more demanding. It often requires a dedicated live-captioning service. For lower-stakes content, a clear statement that a captioned recording will be made available afterwards is a reasonable alternative. If live captioning genuinely isn’t practical for a specific event, publishing an accurately captioned recording promptly afterwards is a reasonable and common approach.

Video that conveys meaning visually: audio description and little dialogue or narration

A video relying heavily on visual demonstration can exclude a blind or low-vision viewer, even with accurate captions. This applies to a video showing a process step by step with minimal spoken dialogue or explanation. It happens because captions only address the audio. An audio description track, or a companion text description of the key visual content, addresses this separately from captioning.

Where this fits into a content and copywriting brief

Deciding what captioning and transcript standard applies to a business’s video content is naturally part of the same content brief that covers written copy, rather than a separate technical afterthought handled only by whoever uploads the video. This is covered more broadly on website copywriting.

Caption formatting, speaker identification and non-speech sound effects, and why readability matters as much as accuracy

Beyond word-for-word accuracy, caption readability depends on reasonable timing. Captions should not flash past faster than a viewer can read. Line breaks need to be sensible, and should not split phrases awkwardly. It also depends on identifying which speaker is talking, when more than one person appears in a video. And it depends on noting non-speech sound effects — a knock at the door, a phone ringing — where the sound carries meaning a viewer without audio would otherwise miss. A technically accurate but poorly formatted caption track can still be difficult to follow. This is particularly true for viewers who rely on captions as their primary way of accessing the content, rather than as a supplementary aid.

Third-party embedded video, and what you don’t control

Video embedded from a third-party platform — a YouTube or Vimeo embed, for instance — inherits whatever captioning that platform’s player provides. This may or may not match the standard your own site otherwise holds itself to. If accessible video is a genuine priority, check specifically what captioning options the embedding platform offers. Confirm whether they’re enabled and accurate. Do not assume an embedded video meets the same standard as content hosted directly.

Why this connects to the legal exposure, not only good practice

Video and audio content that carries meaningful information — a product demonstration, a service explanation, testimonial content used in marketing — falls within the same accessibility obligation as any other content on a site. That obligation is discussed on website accessibility in Australia. Treating video as exempt from accessibility consideration, because it’s “just marketing content” rather than core site functionality, is a common but mistaken assumption.

What to do next

Review and correct any auto-generated captions on existing video content before assuming they’re accurate. Produce transcripts for any audio-only content published on your site. Where this discipline fits alongside the rest of a site’s written content is covered on website copywriting.