Subtitle character counter

The figure comes from the streaming and broadcast style guides, of which the Netflix timed text specification is the most widely copied because it is published openly. It sets 42 characters per line for languages written in Latin script, with a maximum of two lines per caption. Enough studios and vendors adopted it that 42 became the default a freelancer is assumed to be working to unless told otherwise.

It is a default and not a law. Subtitling that inherited its constraints from teletext works to roughly 37 characters, which is why older broadcast material looks more tightly cut. Some vendors permit 47. Languages written in full width characters are counted on an entirely different basis, with Japanese and Chinese guidelines specifying far fewer characters per line because each one occupies the width of two Latin letters. If you are delivering to a client, their specification beats every number on this page.

The limit also exists for a reason worth understanding rather than obeying blindly. Two lines of 42 characters is approximately the amount of text a viewer can absorb without their eyes leaving the picture for long enough to lose the scene. That is the actual goal. A caption that meets the character limit but sits on screen for half a second fails the same test, which is why line length alone is never sufficient.

The same specifications that set 42 characters also set a reading rate, and it is the rate that usually fails first. Netflix specifies 17 characters per second for adult programming and 13 for children's programming, alongside a minimum caption duration of five sixths of a second and a maximum of seven seconds. A caption inside the line limit but shown too briefly is unreadable, and no character counter will tell you so.

The arithmetic is easy to run by hand. Divide the characters in a caption by the number of seconds it stays on screen. A 40 character caption held for two seconds is 20 characters per second, which is past the comfortable ceiling even though the line length passes. The fix is to extend the cue if the audio allows it, or to cut words if it does not.

This is where verbatim transcription and good subtitling part company. Speech that arrives faster than a viewer can read has to be condensed, which means removing filler, contractions of long phrases and repetitions while preserving meaning. It feels wrong the first time you do it. It is standard practice, and the alternative is captions nobody finishes reading. To find out whether your source material is fast enough to force this, our words per minute calculator measures the pace directly from a transcript.

Break at a point where the language already breaks. After a full stop or a comma is best, before a conjunction such as and or but is next, and before a preposition beginning a new phrase works well. The reader should be able to take the first line as a complete unit of sense before the second arrives.

There are splits to avoid regardless of how neat they look. Do not separate a first name from a surname, a title from the name it belongs to, a number from its unit, or an article from the noun it introduces. Each one forces the reader to hold an incomplete fragment while their eyes travel, which is precisely the delay the line limit exists to prevent. Grammar wins over making the two lines the same length.

When splitting is not enough, cut. A line that still exceeds the limit after a sensible break is telling you the caption carries more words than its time on screen can support. Shortening the text is the correct answer rather than shrinking the font or stretching the cue into the next shot.