Stockertim Dictionary · Artificial intelligence
Automatic captioning
Also called auto-captions, automatic captions, auto captioning, machine captioning
In short
Using speech recognition software to produce captions for video or audio, which a person then reviews and corrects before publishing.
- Beginner
- Process
- Checked
In detail
Automatic captioning uses speech recognition to turn the spoken words in a video or audio file into timed captions. The software listens to the soundtrack, predicts the words and splits them into short lines synchronised with the speech. Many editing and publishing tools offer it, and the same transcripts can feed translation and searchable text.
Accuracy depends on audio quality, accents, background noise, overlapping speakers and specialist vocabulary. Even a high word accuracy leaves errors that matter: wrong names, wrong numbers, missing negatives and mistakes in technical terms. Automatic captions also rarely describe important sounds, such as music or a door slamming, which full captions for deaf and hard-of-hearing viewers include.
Captions matter for accessibility, for people watching without sound and for comprehension in a second language. Accessibility guidelines, as published, expect captions for prerecorded video with speech, and accurate captions are part of meeting that expectation. That is why the automatic output is treated as a first draft that a person corrects.
At work, teams generate captions automatically, review them against the audio, fix names, numbers and timing, add sound descriptions and then publish.
Worked example
A 6-minute training video from Lagoon Logistics contains about 900 spoken words.
- Automatic captions are about 95% accurate, so about 5% of words are wrong.
- Words to correct: 900 x 0.05 = 45.
- The reviewer finds most errors in product names and figures, and adds 3 sound descriptions.
- Review time: 6 minutes of video x 3 = 18 minutes.
About 45 errors are fixed in 18 minutes, giving captions that are accurate and accessible.
In practice
Video and content teams use automatic captioning as a first draft, then review every caption file, fix names, numbers and timing, and keep a glossary of product terms to speed up corrections.
Watch out
- Captions include meaningful sounds; a transcript is the text alone.
- A high accuracy figure can still hide errors in the words that matter most, such as names and numbers.
Learn it properly
Official sources
- WCAG standards and guidelines · W3C Web Accessibility Initiative
- Writing for GOV.UK · Government Digital Service
Facts last checked on 3 November 2026.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.