Reading Time Calculator
Reading time with the rate on the outside, code counted separately, images counted at all, and a range instead of a single minute the estimate cannot support.
Words 95
Words in code blocks 8
Code blocks 1
Sentences 5
Characters 606
Rate used 150 words a minute (technical or academic)
Prose 38 sec
Code, at half the rate 6 sec
3 images 33 sec
Reading time 1 min 17 sec
Fair range 58 sec to 1 min 48 sec
What to publish 1 min read
150 words a minute is material the reader stops to think about. The 200
figure everybody quotes comes from silent reading of easy prose on
paper; screens are slower and the range between readers is wider than
the difference between any two published averages, which is why a range
is given here.
1 code block with 8 words in it, counted at half the prose rate. That is
generous: in practice a reader either skips a block or spends a minute
on six lines, and no average covers both.
Images add 33 sec, on the model Medium uses: twelve seconds for the
first image, falling by one per image to a floor of three. It is a
convention rather than a measurement, and it is better than counting
images as nothing.
A published reading time is a promise. Round it to the minute and round
it honestly: a "5 min read" that takes twelve is the last time that
reader trusts the number. If the estimate is over about fifteen minutes,
the useful thing to add is a contents list rather than a bigger number.
The reason to publish one at all is that it measurably improves the
decision to start reading, and the reason to keep it honest is that it
measurably damages trust when it is wrong. Both effects are real and
they point the same way.
Output is valid and updates as you type.
Fix the highlighted fields to update the output.
Reading time is words divided by words per minute. What makes most published reading times wrong is the rate.
200 words a minute is the figure everybody uses. It comes from studies of silent reading of easy prose on paper. Reading on a screen is slower. Technical material with code in it is much slower, because the reader stops. And a post full of screenshots takes time that has no words in it at all.
So this puts the rate on the outside where you can argue with it, counts code separately, counts images at all, and gives a range rather than a single minute the estimate cannot support.
How to use
- Paste the text. Markdown is fine; fenced code blocks are pulled out and counted at half the rate.
- Pick the audience, which is the assumption doing all the work.
- Put in the number of images.
- Publish the rounded figure, and only if you believe it.
Example
A 907-word technical post with one code block and three images:
Words 907
Words in code blocks 7
Code blocks 1
Sentences 2
Characters 4,591
Rate used 150 words a minute (technical or academic)
Prose 6 min 03 sec
Code, at half the rate 6 sec
3 images 33 sec
Reading time 6 min 41 sec
Fair range 5 min 01 sec to 9 min 22 sec
What to publish 7 min readThe same post at 230 words a minute reads as four minutes. Neither number is wrong; the difference between them is entirely the assumption about who is reading, which is why the rate is a field and not a constant.
Pitfalls
A published reading time is a promise. A “5 min read” that takes twelve is the last time that reader trusts the number. Rounding up is the safer error.
Code is not prose. A hundred words of JavaScript is not a hundred words of English: the reader either skips it entirely or spends a minute on six lines. Counting it at half the prose rate is generous and it is better than averaging it in.
Images take time. Medium’s model, twelve seconds for the first image falling by one to a floor of three, is a convention rather than a measurement, and it is a great deal closer than counting them as zero.
The range between readers is wider than the difference between any two averages. A slow reader and a fast reader differ by a factor of two on the same text. That is why the range is printed, and why arguing about 200 against 230 words a minute is not worth the time.
Reading rate is not comprehension. Skimming at 400 words a minute is reading in the sense that the eyes moved. If the post is a tutorial, the useful estimate is how long it takes to follow, which is longer than any of these figures.
A long estimate is a signal, not a badge. Past about fifteen minutes, the thing to add is a contents list and some subheadings rather than a bigger number.
Non-native readers are a large share of any English-language audience. The second-language rate is in here because leaving it out is a decision too, and it is the rate closest to a global audience.
Compatibility
Everything runs in the browser: nothing is uploaded and nothing is stored.
Words are counted with Unicode property escapes, so accented and non-Latin text counts correctly. A hyphenated word counts once, because that is how it is read, and an apostrophe is part of the word. Fenced code blocks in both the backtick and tilde forms are removed before the prose count and counted separately.
The rates are the middle of the published range for each case rather than one study’s figure: 230 for general adult silent reading, 150 for material the reader stops to think about and for reading aloud, which is also the rate audiobooks are timed at, 400 for skimming and 130 for fluent second-language reading. Every one of them is an average with a wide spread.
What this does not do is estimate difficulty from the text itself. Sentence length and syllable counts give a readability grade, which is a different measurement with its own caveats; the content score checker on this site does that, and combining the two into one number would hide both.