Captions must be read while the viewer is also watching the picture. A structurally valid file can still flash past too quickly or wrap into a block that hides the demonstration. Treat the checker as a list of places to review, not a certificate of accessibility.
Know what the number measures
Characters per second compares the visible text length with the time a cue stays on screen. A dense cue or a short duration raises it. It does not measure a viewer’s comprehension, familiarity with the topic or ability to read while following the action.
Netflix's English style guide allows 42 characters on a line and at most two lines.
Netflix's English style guide sets reading speed at up to 20 characters a second for adult programmes and up to 17 for children's.
Those are delivery guidelines for that context, not a universal rule for every audience. Use the checker’s thresholds as an explicit review policy. Consider children, unfamiliar names, technical vocabulary and the amount of visual attention the video requires.
Free tools for this
Free to use, no account needed.
- Subtitle readability checkerCheck a caption file for lines that are too long, too fast, too short or overlapping, and fix them.
- Subtitle editorFix caption text and timing line by line, with your video playing alongside.
- Add subtitles to videoBurn an SRT or WebVTT caption file into a video so the captions show on every platform.
Review flagged cues against the recording
Do not delete qualifications, negations or useful names to make a cue look shorter. If verbatim speech cannot fit comfortably, choose a less crowded passage or make a deliberate editorial decision about the presentation. Do not pass off a paraphrase as a verbatim transcript.
- Load the complete caption file in the checker. Read the flags for long lines, fast reading, overlaps and invalid durations separately.
- Choose a flagged cue and listen to its passage in the original. Establish whether the text is accurate before changing it to fit a threshold.
- Break a crowded cue at a natural phrase boundary. Keep words together where splitting them would change meaning or force the viewer to wait for the object of a sentence.
- Adjust cue times only within the speech and context they represent. Do not extend a caption over unrelated dialogue just to lower its reading-speed number.
- Download the revised file, rerun the check and watch it on the final video. Inspect the saved captions rather than only the editor fields.
Look beyond the threshold
A silent interval may carry important sound information even when nobody speaks. Decide whether the publication needs descriptions of meaningful sounds. A spoken-word transcript alone does not cover every accessibility need.
- Read captions at the intended display size. A line that fits technically may still be too small or sit over a distracting background.
- Check speaker changes. Make attribution clear when the picture does not show who is talking.
- Check names, units, numbers and specialist vocabulary against a spelling list or the original source.
- Watch the picture with captions enabled. Keep essential gestures and labels visible.
- Check intentional overlaps manually. Simultaneous speech is an editorial problem, not automatically a malformed file.
Approve the exact publication version
Burned-in captions cannot be switched off or corrected separately from the video. Check the layout before burning them, then watch the exported picture to confirm the renderer used the intended position and size.
A separate caption file gives the player control over display. Its wrapping may differ from your preview. Test it in the publication system and confirm the language is identified correctly.
Keep the approved caption file with the approved video version. Cutting the video after approval can invalidate timing even if none of the caption words changes. Repeat synchronization checks after timeline changes.
Questions about this workflow
Does a clean report mean the captions are correct?
No. It means the file passed the implemented structural and reading checks. Spelling, meaning, speaker attribution and synchronization still need review.
Should I remove words from a fast cue?
Not automatically. Try phrase-aware splitting and appropriate timing first. Do not remove a qualification or negation to satisfy a number.
Why does the player wrap lines differently?
Its font, size and available frame determine wrapping. Check the receiving player, not just the text file.
Sources and further reading
Want captions without the cleanup?
Studio captions every word you said, a few words at a time, and changes any line you ask about in plain words.
