Caption Readability Checker

Knowledge Base

What This Tool Checks

Six things, per cue: reading speed in characters per second (the single best readability predictor. A caption that flashes by unread might as well not exist), line length, lines per cue, minimum and maximum duration, the gap between consecutive cues (captions that butt against each other read as one flickering block), and overlapping timings, which are almost always an export error. Speaker labels and sound-effect tags count toward the character totals because viewers read them too.

The Defaults Are Conventions, Not Law

The preset numbers (17 CPS, 42 characters per line, two lines, 1 to 7 second durations) follow widely used professional subtitling conventions of the kind published in broadcaster and streaming-platform style guides. But every client and platform publishes its own numbers, they differ (children's content, for instance, typically uses much slower reading speeds), and they change. If you're delivering against a spec, check the current version of that spec and set the fields to match. The defaults are for creators who don't have one and want a sane baseline.

Fixing What It Flags

High CPS has two fixes: extend the cue's duration if there's room before the next one, or, usually better, trim the text. Captioning is allowed to lightly condense. Verbatim transcription that nobody can read fast enough serves no one. Line-length violations usually want a line break moved at a natural phrase boundary rather than mid-clause. To apply fixes, round-trip the file through the SRT Caption Editor, then re-check here. For the reasoning behind these guidelines, see the caption quality guide.

FAQ

Q: Where do the default thresholds come from?
A: They follow widely used professional subtitling conventions: roughly 17 characters per second reading speed, 42 characters per line, two lines per cue, and cue durations between about 1 and 7 seconds. They are conventions, not law: broadcasters and platforms publish their own style guides with different numbers, so if you're delivering to a specific client or platform, check their current spec and adjust the fields to match.

Q: Is my caption file uploaded anywhere?
A: No. The file is parsed and analyzed entirely in your browser with JavaScript: nothing is sent to a server, and the page works offline once loaded.

What to do when a check fails

A reading speed failure has two fixes and they are not equivalent. Extending the cue's duration works where there is room before the next one and where the picture allows. Splitting the cue into two works everywhere but needs a sensible break point. Reducing the text is the last resort, because captions should carry what was actually said rather than a summary of it.

Line length failures are usually solved by breaking differently rather than by cutting words. Breaking at a grammatical boundary, before a preposition or after a complete clause, keeps a line under the limit while reading naturally. Breaking wherever the character count runs out produces lines that are technically compliant and awkward to read.

A cue that is too short is nearly always a timing artefact rather than a content decision. Automatic transcription produces these constantly, splitting a phrase across two cues a fraction of a second apart. Merging them is almost always correct, and doing it in bulk is faster than reviewing each one.

Overlaps need resolving in one direction consistently. When two cues overlap, either trim the first or delay the second, and picking one approach for the whole file keeps the result predictable. Alternating between the two produces a file where timing relationships to the picture vary unpredictably.

Re-run the check after fixing, because fixes create new failures. Extending a cue to solve a reading speed problem can create an overlap with the next one, and splitting a cue can produce a fragment below minimum duration. One pass is rarely enough, and two or three iterations to a clean result is normal rather than a sign of a bad file.

Vocabulary

Characters per second. A measure of how fast caption text arrives, used alongside words per minute. Some specifications state limits in one, some in the other, and a file can pass one while failing the other.

Cue merge. Combining two adjacent cues into one. The usual fix for fragments produced by automatic transcription, where a single phrase gets split across cues too brief to read.

Break point. The place a caption line wraps. Choosing a grammatical boundary rather than the character limit keeps lines readable, and is the difference between compliant captions and comfortable ones.

Iteration. Running a check again after applying fixes, because corrections frequently create new problems. Extending one cue can overlap the next, so reaching a clean result usually takes several passes.