Editing Automatic Captions for Accuracy
Editing Automatic Captions for Accuracy records a distinct W3C-supported caption-writing task while keeping implementation mechanics and unsupported house-style rules outside the guidance.
Automatically generated captions should be reviewed and edited for accuracy before they are treated as sufficient captions.
Key details
Important caveats
Scope Boundary
Do not assume an automatic caption track is adequate merely because one exists; W3C notes that automatic captions are not sufficient unless confirmed to be fully accurate.
Further guidance
Content
Use automatic speech recognition as a starting point, then correct wording, omissions, punctuation, timing, speaker information, and other meaning-bearing errors as needed.
Purpose
Editing prevents recognition mistakes from changing or contradicting the spoken meaning.
Sources and evidence
Sources are shown with the role they play in this guide. Historical or style-sensitive claims are kept within the evidence boundary described above.
- Captions/Subtitles — W3C WAI (opens in a new tab)World Wide Web Consortium (W3C) Web Accessibility Initiative · W3C WAI guidance that captions are synchronized text alternatives for speech and important audio information, with timing encoded in caption files.
- Transcribing Audio to Text — W3C WAI (opens in a new tab)World Wide Web Consortium (W3C) Web Accessibility Initiative · W3C WAI guidance for identifying speakers and representing relevant audio information in transcripts and captions.