Your organization publishes captions with every video. You've checked the compliance box. But when a Deaf employee or customer reviews that content, can they actually follow the conversation? Can they identify who's speaking during a panel discussion? Do they know when an alarm sounds or when the audience laughs?
WCAG Success Criterion 1.2.2 requires captions for prerecorded video, but the standard goes further than synchronized text. The Understanding document specifies that captions must include dialogue, identify speakers, and convey non-speech information through sound. Most caption audits stop Assistive Technology checking whether a caption file exists. This checklist evaluates whether your captions deliver an equivalent experience.
Prerequisites
Before using this checklist, confirm:
- You have access to the video player your audience will use (not just the authoring environment).
- You can review caption files in their native format (.srt, .vtt, or platform-specific).
- You have documentation of your caption production method (automated tool, vendor, in-house captioner).
- You can test with captions enabled in the delivery environment.
Caption Quality Checklist
1. Dialogue Accuracy
Check: Play a 2-minute segment and compare spoken words to caption text. Count errors.
Requirement: Zero substantive errors. Filler words ("um", "uh") may be omitted if meaning is preserved, but self-corrections that clarify intent must appear ("slide five... I mean slide six" becomes "slide six").
Good looks like: A Deaf viewer receives the same factual content as a hearing viewer. Technical terms, product names, and numeric values appear exactly as spoken.
2. Speaker Identification
Check: Review any segment where multiple speakers appear. Verify that each speaker change includes an identifier in the caption.
Requirement: Captions must identify the speaker when the speaker changes, even if the change is visually obvious. Use consistent labels (e.g., "Dr. Martinez:", "Interviewer:", "Audience Member:").
Good looks like: A viewer who looks away from the screen can return and immediately understand who is speaking without rewinding.
3. Non-Speech Audio
Check: Note every meaningful sound in a 5-minute segment: door slams, phone rings, applause, background music that sets mood, off-screen voices. Verify each appears in brackets or parentheses.
Requirement: Any sound that contributes to understanding must be captioned. Format: [phone ringing], (applause), [door slams].
Good looks like: A viewer understands why a character reacts to something off-screen or knows that a warning alarm has sounded.
4. Caption Timing
Check: Identify the longest caption in your video. Measure how long it remains on screen. Count words.
Requirement: Captions must remain on screen long enough to be read. A general guideline: 160-180 words per minute for adult readers. Captions should appear when the audio begins and disappear when it ends, with minimal lag.
Good looks like: Viewers can read the caption without pausing the video. No caption disappears before a viewer can finish reading it.
5. Line Length and Break Points
Check: Review 10 random captions. Count characters per line. Note where line breaks occur.
Requirement: Maximum 32-42 characters per line (consult BBC Subtitle Guidelines or DCMP Captioning Key for platform-specific limits). Break lines Assistive Technology natural phrase boundaries, not mid-word or mid-clause.
Good looks like: "The compliance officer reviews / the accessibility statement" (not "The compliance officer / reviews the accessibility statement").
6. Caption Placement
Check: Identify any captions that obscure on-screen text, speaker faces, or critical visual information.
Requirement: Captions should appear in a consistent location (typically bottom-center) and not block important content. If on-screen text appears, reposition captions temporarily.
Good looks like: A viewer can read both the caption and any on-screen text, charts, or names simultaneously.
7. Formatting Consistency
Check: Review how italics, speaker labels, and sound effects appear across the entire video. Note any formatting inconsistencies.
Requirement: Use consistent formatting rules. Example: speaker labels in all caps before dialogue, sound effects in brackets, off-screen dialogue in italics.
Good looks like: A viewer learns your formatting system within 30 seconds and applies it throughout the video without confusion.
8. Caption File Accessibility
Check: Verify that captions can be toggled on/off (closed captions) rather than burned into the video (open captions), unless there's a specific reason for open captions.
Requirement: Closed captions allow viewers to customize appearance (font size, color, background) based on their needs. Open captions are acceptable for social media previews or when platform limitations require them.
Good looks like: A viewer with low vision can increase caption font size without requesting a separate video file.
Common Mistakes
Trusting the production method over the output. Automated tools have improved, but they rarely identify speakers or capture non-speech audio. Professional vendors sometimes deliver captions worse than automated output. Always audit the final result.
Treating verbatim as a quality marker. Including every "um" and stutter doesn't improve accessibility. Captions should convey meaning, not transcribe noise.
Skipping the delivery environment test. Captions that look perfect in your authoring tool may have timing issues, placement problems, or formatting errors in the player your audience uses.
Assuming visual clarity eliminates the need for speaker identification. A viewer who glances away, uses a screen magnifier, or has low vision may not catch visual speaker changes.
Next Steps
Run this checklist on three representative videos from your library: one with automated captions, one with vendor-supplied captions, and one with in-house captions. Document the failure rate for each production method.
If automated captions fail items 2, 3, or 5, budget for human review or a hybrid approach. If vendor captions fail any item, send them this checklist as part of your quality standard and request remediation.
Caption quality isn't binary. The question isn't whether captions exist but whether they deliver the equivalent experience that WCAG requires. A Deaf compliance officer reviewing your training videos should receive the same information, Assistive Technology the same pace, with the same context as their hearing colleagues. Anything less is partial conformance.
