Subtitle reading speed standards: the numbers Netflix, the BBC, and WCAG actually use

The professional guides disagree on almost nothing here. Adult subtitles should not run faster than roughly 160-180 words per minute, which works out to about 15-20 characters per second. Netflix, the BBC, and the US Described and Captioned Media Program all land in that band, and they have converged on it for the same reason: past that speed the average viewer stops finishing lines before they disappear.

Two things worth saying up front. These are comfort ceilings, not laws of physics, and they shift with the audience: children, older adults, and people reading in a second language read slower, so their ceilings are set lower. And the one standard people most often name as the source of these numbers, WCAG, actually specifies none of them. The figures below come from broadcasters and captioning bodies, not from the accessibility spec.

Netflix publishes its rules openly, which is why they have become the de facto reference for a lot of the industry. The English (USA) Timed Text Style Guide sets reading speed at up to 20 characters per second for adult programs and up to 17 CPS for children's programs. Lines are capped at 42 characters, with a maximum of two lines per subtitle. (Last verified: 2026-07-13.)

Netflix also fences in the timing at both ends. Each subtitle event must stay on screen for a minimum of 20 frames (five-sixths of a second, about 0.83 seconds at 24fps) and a maximum of 7 seconds. The floor stops a caption from flashing; the ceiling stops a single line from lingering long after the speaker has moved on.

The BBC Subtitle Guidelines frame the same idea in words per minute rather than characters. The recommended reading speed for a general adult audience is 160-180 wpm, which is roughly 0.33 to 0.375 seconds of screen time per word. Read the other way, that sets a minimum duration: a subtitle should stay up for about 0.3 seconds per word, so a four-word subtitle needs roughly 1.2 seconds on screen before it can leave.

The BBC's line rules are shaped by where the text appears. Broadcast (teletext) subtitles are limited to 37 fixed-width characters per line, a constraint of the old monospaced grid rather than a universal rule. For online video the recommendation is width-based instead: a maximum line length of about 68% of a 16:9 frame (90% for 4:3). The guidance is a maximum of two lines per subtitle for landscape and square video, rising to three lines for 9:16 vertical.

The Described and Captioned Media Program writes captions for classrooms, and its Captioning Key ceilings are deliberately more conservative because the audience is still learning to read. The presentation rate is capped by educational level: not to exceed 130 wpm at the lower level, 140 wpm at the middle level, and 160 wpm at the upper level.

The Captioning Key also sets duration bounds: a minimum caption length of 40 frames (about 1.33 seconds at 30fps) and a maximum of 6 seconds, with a preference for no more than two lines per caption. The pattern across all three guides is the same shape – a floor so text does not flash, a ceiling so it does not overstay – with the numbers tuned to who is reading.

Here is the correction that matters most. WCAG Success Criterion 1.2.2 (Captions, Prerecorded) requires that captions exist and are synchronized with the audio, but it specifies no reading speed at all – no words per minute, no characters per second, no minimum display duration. The success criterion is written to be technology- and content-agnostic on purpose.

So when a secondary blog tells you that "WCAG requires 160-180 wpm" or "WCAG allows 20 CPS," it is quietly borrowing a broadcaster convention and stamping the wrong name on it. Those numbers are real and useful, but they come from the BBC, Netflix, and the DCMP. If you need a defensible reading-speed target for an accessibility audit, cite the caption body you are following, not the accessibility spec.

The obvious question is why 180 wpm is treated as fast when people plainly read faster than that. A 2019 meta-analysis of 190 studies covering 18,573 participants put average adult silent reading at 238 wpm for non-fiction and 260 wpm for fiction, with oral reading around 183 wpm. On paper, a reader has plenty of headroom above the caption ceilings.

The gap exists because reading a subtitle is not the same task as reading a page. On screen the viewer is splitting attention: following the picture, tracking who is speaking, and reading the line, all at once. The caption ceilings are set below the silent-reading rate to leave room for everything else the eye has to do. That is also the honest reason word counts feel slippery on video – if you have ever tried to estimate them, our note on how many words are in an hour of audio explains why speech and reading rates rarely line up.

Because the guides mix units, it helps to be able to move between them. As a rough English conversion, 180 words per minute corresponds to about 15 characters per second. That is why the BBC's 180 wpm and Netflix's up-to-20 CPS are broadly comparable ceilings rather than contradictory ones – they are the same target expressed in different currencies, with Netflix leaving a little more headroom.

The conversion is only a guide, because it depends on average word length, and average word length depends on language. It also depends on the reader. The reason children's captions are pegged lower – 17 CPS for Netflix kids' content, 130 wpm at the DCMP's lower level – is that the meta-analysis is clear that children, older adults, and second-language readers all read more slowly. Treat every number here as a comfort ceiling for its stated audience, not a universal hard limit.

If you are captioning your own video, these standards turn into a two-step job. Step one is an accurate transcript: the actual words, correctly spelled, with usable timestamps. Step two is timing that transcript to a reading speed - splitting long sentences, holding each cue for its minimum, and keeping the CPS under the ceiling for your audience. A subtitle generator or a video-to-text pass gets you the raw SRT; the reading-speed pass is where you apply the numbers above. If accuracy is your worry before you even reach the timing, caption accuracy standards covers what "accurate enough" means for captions, and transcription vs translation matters the moment your subtitles cross a language.

This is where Pepys fits, and only here. Pepys is not a study author and did not set any of these ceilings - it produces the honest raw transcript and SRT file that you then time to the BBC, Netflix, or DCMP standard your project needs. The numbers are the industry's; the clean starting file is the part we can hand you.