Give the viewer time to read.

What CPS measures

Characters per second is visible character count divided by cue duration. In Arbray, spaces count, line breaks do not, and text markup is excluded from the count. Unicode code points are counted, so a combined emoji or accented sequence can contain more than one counted unit. The number is a repeatable screening metric, not an exact model of human reading.

A 48-character caption shown for two seconds has a rate of 24 CPS. Extending it to three seconds reduces the rate to 16 CPS, but that only helps if the extra time fits the speech and the next caption. Slowing text by creating a new overlap trades one problem for another.

CPS = visible characters ÷ duration in seconds
48 ÷ 2 = 24 CPS
48 ÷ 3 = 16 CPS

Set targets for the audience

The workbench starts with 20 characters per second and 42 characters per line. These are editable review targets selected for this tool, not universal accessibility limits or a platform-approved preset. Your client, language, viewer age and delivery specification may require other values.

Read every fast caption in context. Options include simplifying text, splitting a long cue, moving a boundary where the speech allows it or leaving a deliberate fast passage unchanged. This tool does not rewrite meaning or automatically extend every caption.

Line length is only one part of layout

A long line may be hard to scan, but an awkward short line can be hard to read too. Arbray wraps plain text at word boundaries using your line-length target. It does not split a long word or guess linguistic phrase boundaries, and it leaves tagged text unchanged to avoid altering styling scope.

Automatic wrapping can create more than two lines. That remains visible as a finding; use the cue editor to choose a better break or shorten the wording. Actual width depends on the player, font and screen. Character count cannot guarantee that text fits a phone.

Do a final viewing pass

Play the exported subtitles with the video. Check that each caption stays long enough, speaker changes are understandable and useful non-speech information is present when needed. A subtitle file alone cannot tell Arbray whether an alarm, a name or a key line of dialogue is missing. Technical checks support the editorial pass; they do not replace it.