# Captions and Transcripts

Captions synchronize speech and meaningful audio with media, while transcripts provide a separate text representation that can be read and searched independently.

- Type: artifact
- Domain: Experience & interface
- Level: Foundational
- Last reviewed: 2026-09-24
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/captions-and-transcripts/>)

## What it means

Captions present dialogue, speaker identification, and relevant non-speech sound in time with audio or video. A transcript provides the media's meaningful spoken and audio information as text, sometimes adding descriptions of important visual content. Neither is automatically a complete substitute for the other.

## Why it matters

Captions give Deaf and hard-of-hearing users access during playback and help people in noisy, quiet, or unfamiliar-language contexts. Transcripts support reading, search, quotation, translation, and alternative pacing.

## When to use it

Provide accurate captions for prerecorded video with audio and for live media where required; provide transcripts for audio-only content and whenever independent text access is useful. Include visual descriptions in a transcript when visuals convey information not available through audio.

## How it works

1. Create a verbatim-enough text record that preserves meaning, speakers, and relevant sounds.
2. Edit automated output for names, terminology, punctuation, timing, and omissions.
3. Synchronize readable caption segments with the media and identify speakers when needed.
4. Publish a logically structured transcript near the media and include necessary visual descriptions.
5. Test caption controls, resizing, contrast, synchronization, mobile playback, and transcript navigation.

## A practical example

A product webinar includes edited closed captions identifying each speaker and noting meaningful demo sounds. A linked transcript uses headings for sections and describes visual changes in the prototype, making the session searchable and usable without playback.

## Where it breaks

Automated speech recognition requires human review and performs poorly with specialized terms, overlapping speakers, and noisy audio. Live captions may lag, and transcripts without visual descriptions can omit essential demonstration content.

## Common mistakes

- Publishing unreviewed automatic captions as final.
- Omitting speaker changes and meaningful non-speech audio.
- Treating a transcript as sufficient for synchronized video captions.
- Burning captions into video with no way to customize their presentation.

## Visualization

[View the visualization on the website](<https://wiki.soroush.uk/topic/captions-and-transcripts/#visualization-captions-and-transcripts>)

## Sources

1. [Captions/Subtitles](<https://www.w3.org/WAI/media/av/captions/>) — W3C Web Accessibility Initiative · Accessibility guidance
2. [Transcripts](<https://www.w3.org/WAI/media/av/transcripts/>) — W3C Web Accessibility Initiative · Accessibility guidance
3. [Understanding Success Criterion 1.2.2: Captions \(Prerecorded\)](<https://www.w3.org/WAI/WCAG22/Understanding/captions-prerecorded.html>) — W3C Web Accessibility Initiative · Standards guidance

## Related topics

- [Accessibility](<https://wiki.soroush.uk/topic/accessibility/>)
- [Content Design](<https://wiki.soroush.uk/topic/content-design/>)
- [Plain Language](<https://wiki.soroush.uk/topic/plain-language/>)
